The Intonation of Interrogation in Palermo Italian: Implications for intonation theory 3484303344, 9783484303348

Citation preview

Linguistische Arbeiten

334

Herausgegeben von Hans Altmann, Peter Blumenthal, Herbert E. Brekle, Gerhard Heibig, Hans Jürgen Heringer, Heinz Vater und Richard Wiese

Martine Grice

The intonation of interrogation in Palermo Italian Implications for intonation theory

Max Niemeyer Verlag Tübingen 1995

Die Deutsche Bibliothek - CIP-Einheitsaufhahme Grice, Martine: The intonation of interrogation in Palermo Italian: implications for intonation theory / Martine Grice. - Tübingen: Niemeyer, 1995 (Linguistische Arbeiten ; 334) NE: GT ISBN 3-484-30334-4

ISSN 0344-6727

© Max Niemeyer Verlag GmbH & Co. KG, Tübingen 1995 Das Werk einschließlich aller seiner Teile ist urheberrechtlich geschützt. Jede Verwertung außerhalb der engen Grenzen des Urheberrechtsgesetzes ist ohne Zustimmung des Verlages unzulässig und strafbar. Das gilt insbesondere für Vervielfältigungen, Übersetzungen, Mikroverfilmungen und die Einspeicherung und Verarbeitung in elektronischen Systemen. Printed in Germany. Druck: Weihert-Druck GmbH, Darmstadt Einband: Hugo Nädele, Nehren

Contents

Acknowledgements

xi

Chapter 1 - Introduction 1 Standpoint and objectives. 2 The concept of the nucleus. 3 Chapter outline.

1 1 2 3

Chapter 2 - The British nucleus-plus-head approach. 1 Introduction. 2 The tone unit. 3 Componentiality 3.1 The prehead-head distinction. 3.2 The nucleus-tail distinction. 3.2.1 The nucleus, the nuclear syllable and the nuclear tone. 3.2.2 The domain of the nuclear tone. 3.2.3 Couper-Kuhlen: Steep slope, gradual slope and flattening 3.2.3.1 Steep slope. 3.2.3.2 Gradual slope 3.2.3.3 Flattening 3.2.4 The interdependence of nuclear syllable and tail 3.3 The head-nuclear syllable transition 4 A working model

7 7 8 9 10 11

16 16 17 18 19 20 21

Chapter 3 - A British-style analysis of Palermo Italian. 1 Introduction 1.1 Corpus details. 2 Transcription 3 Formal analysis of the core corpus 3.1 Nuclear tones. 3.1.1 Simple tones. 3.1.1.1 Falling tones.

23 23 24 25 25 25 26 26

12 13

vi 3.1.1.2 Level tones 3.1.1.3 Rising tones 3.1.2 Complex tones. 3.2 Types of prehead. 3.3 Types of head. 3.3.1 Falling 1 3.3.2 Falling 2. 3.3.3 Falling 3. 3.3.4 Falling 4. 3.3.5 Falling 5. 3.3.6 Low. 3.3.7 High 3.3.8 Mid. 3.3.9 Rising 3.4 Frequency of occurrence of prenuclear contours. 4 Functional analysis of the core corpus. 4.1 Questions and statements. 4.1.1 Complex tones. 4.2 Lists. 5 Main observations and conclusion.

Chapter 4 - Auditory studies of Italian intonation. 1 Introduction. 1.1 Varieties of Italian. 1.2 Auditory treatments of Italian intonation 2 Agard and Di Pietro. 2.1 Descriptive framework. 2.2 Basic patterns 2.3 Location of distinctive pitch 3 Chapallaz. 3.1 Descriptive framework. 3.2 Basic patterns. 3.3 Location of distinctive pitch 4 D'Eugenio 4.1 Descriptive framework. 4.2 Basic patterns 4.3 Location of distinctive pitch 5 Canepari. 5.1 Descriptive framework. 5.2 Basic patterns 5.3 Location of distinctive pitch

27 28 29 29 29 30 30 30 30 31 31 31 31 32 33 33 33 34 35 35

37 37 37 38 39 39 40 41 41 42 42 44 47 47 48 49 50 50 51 52

vii 5.4 Canepari's account of "Sicilian Italian" 6 Fogarasi. 7 Conclusions 7.1 BP 1 7.2 BPH. 7.3 BP IH 7.4 Distinctive pitch 7.5 Relation to analysis of Palermo Italian. 7.6 Final remarks.

Chapter 5 - Association in intonation models. 1 Autosegmental association 1.1 The theory. 1.2 Universal and language-specific phenomena. 1.3 Downstep 2 Post-autosegmental work on association. 2.1 Association at different levels 2.1.1 Pitch Accents. 2.1.2 Downstep 2.1.3 Levels above the pitch accent. 2.2 Specific models. 2.2.1 Pierrehumbert (1980) 2.2.2 Ladd (1983) and subsequent work. 2.2.3 Hirst (1983, 1986, 1988) 2.2.4 Bruce (1977, 1983, 1987) 2.2.5 Beckman and Pierrehumbert (1986) and Pierrehumbert and Beckman (1988) 3 Concluding remarks

Chapter 6 - Alignment 1 Kingdon. 1.1 (A) Complex tones. 1.2 (B) Contrastive and homosyllabic preheads. 1.3 Comparison of two types of delay: (A) and (B). 2 Autosegmental pitch accent theories. 2.1 Complex tone 2.1.1 Ladd 2.1.2 Pierrehumbert and colleagues 2.2 "Homosyllabic preheads" and other phenomena. 2.2.1 Ladd

52 53 54 54 54 54 55 55 55

57 57 57 62 64 65 65 66 66 66 66 66 69 74 78 82 89

93 94 94 95 98 99 100 100 102 104 104

viii 2.2.2 Pierrehumbert and colleagues 3 The domain of the nucleus. 3.1 Introduction 3.2 Downstep 3.2.1 Distribution. 3.2.2 Shape of downstep. 3.2.2.1 Calling contours. 3.3 The early peak contour. 4 Alternative pitch accent analysis. 4.1 Truly low L tone in H+L* 4.2 Downstep 4.3 Calling contours 4.4 Leading and trailing tones - their differing status. 5 Representation of English pitch accents with the extended structure. 6 Validation of the new pitch accent inventory and structure. 6.1 L-L* pitch accent. 6.2 Neutralisation. 6.3 Downstep in only the second tone of a bitonal pitch accent - implications. 6.4 Validation - conclusion 7 Conclusion

Chapter 7 - Reanalysis of interrogative intonation in Palermo Italian 1 Precursors to the analysis. 2 Implications for the analysis of Palermo Italian. 2.1 Type of utterance needed 2.2 Elicitation method. 2.3 Sentence material 2.4 Speakers. 3 Description of "nuclear" contours found in the corpus. 3.1 Yes-no interrogatives 3.1.1 Late focus. 3.1.2 Early focus. 3.2 Non-final items in lists. 3.3 Declaratives. 3.3.1 Late focus. 3.3.2 Early focus. 3.4 A note on the "head" contours

106 109 109 110 110 110 113 114 117 117 117 118 120 124 125 133 133 135 135 136

139 139 140 140 141 141 144 144 144 144 145 146 147 147 148 171

ix 4 Analysis of "nuclear" intonation patterns 4.1 Interrogative contours. 4.1.1 Number of structural positions. 4.1.2 Type of structural position 4.2 Non-interrogative utterance types. 4.2.1 "Listing" contours. 4.2.2 Declaratives. 5 The prosodic tree in Palermo Italian. 5.1 Central association 5.2 Peripheral association 5.3 Secondary association 5.3.1 Additional evidence for secondary association 6 Timing considerations. 7 How is interrogation signalled? 7.1 Location of the interrogative marker on the tone tier. 7.2 Association of the interrogative marker on the prosodic tier. 7.3 Pitch accent structure. 8 Relation to auditory analysis in Chapter 3 9 Analysis of contours - summary

171 171 172 173 176 176 178 180 182 183 184 186 187 191 191 193 198 200 204

Chapter 8 - Summary and Conclusion. 1 Summary 2 Conclusion

207 207 209

Bibliography.

215

Acknowledgements

I am grateful to David Crystal for supervising my first auditory analysis of Palermo Italian at the University of Reading and, in chronological order, to Giulio Lepschy, Andy Butcher, Linda Shockey and Jill House for supervision during the writing of my doctoral thesis, awarded at University College London, of which this book is a mimimally revised version. During the past year I have continued to work on some of the theoretical issues addressed here. My recent work is to appear in Volume 12.2 of Phonology as "Leading tones and Downstep in English". The book retains the original analysis. Thank you to Carlos Gussenhoven, Frans Hinskens, Mike Johnson and Jacques Koreman who read and commented on early drafts of chapters of the thesis, and to Cinzia Avesani, Bill Barry, Mary Beckman, Daniel Hirst, Bob Ladd and Irene Vogel, who kindly gave me the opportunity to discuss aspects of the work reported here. The revised version has benefitted from extensive comments by Bob Ladd and from discussing with a number of his students at the University of Edinburgh, where the thesis was used as material for a course on intonation. It has also gained from discussion with Mary Beckman, Carlos Gussenhoven and Mike Johnson and participants of the ToBI workshop in Ohio in June 1993. I would also like to thank my colleagues at the University of the Saarland for providing such a stimulating forum for discussion on issues intonational, especially Bistra Andreeva, Bill Barry, Ralf Benzmuller, Jacques Koreman and Michelina Savino. A special thanks goes to Michelina Savino and Bistra Andreeva for invaluable help in preparing the final manuscript. For support on every level I thank Mike Johnson to whom this book is dedicated.

I. Introduction

1 Standpoint and objectives The development of a theory of intonation often involves recourse to a body of data comprising a set of contours used more or less consistently by a group of speakers. A phonological analysis of such a body of data might treat it as a manifestation of an independent system, and adopt a method of analysis which follows the phonemic principle, viz. one which captures all the functional distinctions within that system whilst ignoring any formal characteristics which have little or no functional value. However, if the aim of intonational research is to discover facts about intonation as a universal system, it is worthwhile to develop a system of primitives which accounts adequately for the intonation patterns used in a number of different languages. In both cases, the forms described must be perceivable by ear if they are considered to be potential linguistically perceivable categories. For this reason, consideration is given to auditory impressions as well as to fundamental frequency records extracted by machine. One of the aims of this thesis is to provide taxonomic information on features of Palermo Italian. This is not only so that an analysis of this particular variety can be performed, but also so that a new body of data can be made available as a testing ground for theories of intonation which have been derived from the analysis of other systems. It is also deemed necessary to take another look at a number of accounts of English intonation (developed by, amongst others, Pierrehumbert and Beckman, and Ladd) and to investigate their advantages and shortcomings regarding the analysis of (i) the contours currently described in the literature, and (ii) a number of contours which have been referred to in past work and have since been largely ignored. The main purpose here is to build on the foundations laid by the above authors, in order to provide a framework of intonational analysis which is flexible enough to account for both Palermo Italian and (at least British RP) English. Optimally, such a framework should at once provide for a phonological analysis of the intonation of each language as a system, and shed light on the nature of the differences between the two systems. In this way, the above mentioned universal aim would be adhered to. To the author's knowledge, there has so far been no published analysis of the intonation of Italian spoken in Palermo. As a result, a number of corpora have been drawn up to form the basis of the descriptive and theoretical parts of the thesis. These corpora involve a relatively homogeneous group of speakers, all born and currently living in Palermo, of middle class and with a university or equivalent further education. Such a homogeneous group was chosen because, even within Palermo itself, there is a considerable degree of geographical and sociological accentual variation. It is the theoretical aims stated above which motivate the concentration in the investigations on a set of contours used consistently by a group of informants, rather than on data obtained from an exhaustive sociolinguistic survey.

2 According to Lindsey (1985:3), intonation is "no more than the intersection of a set of phonological (specifically tonal) structures and a set of semantic-pragmatic functions (here termed p-functions)". Here, as in Lindsey's study, the p-function under investigation is interrogation, specifically as it is manifested in polar (yes-no) questions. Whilst English has the choice of using either tonal or other morpho-syntactic means to realise this p-function, Italian relies solely on tonal means. This makes it possible to perform a more consistent analysis of the realisation of this function. For instance, the phrase "Glielo porta domani" may be used as an information-seeking question ("Is she bringing it to him tomorrow?"), or as a statement of fact ("She's bringing it to him tomorrow."); their differentiation relies on the intonation contour. Provided with the opportunity straightforwardly to investigate the intonation contour whilst keeping segmental factors constant, the majority of Italian analysts have taken the distinction between statements and polar questions as a starting point, a lead which will be followed in the present study. Palermo Italian differs in a number of ways from what has been described in text books as Standard Italian, specifically with regard to its manifestation of the interrogative p-function; it exhibits a terminal fall in polar questions which in Standard Italian have terminal rises. However, this terminal fall is preceeded by a rise, and, in certain contextually-determined situations, the terminal fall is altogether absent. It will be shown that, alongside Standard Italian varieties, as well as alongside English, it is the rising element which signals interrogation. The ability to formally account for the contextually-dependent absence of the terminal fall is one of the challenges a theory of intonation must be able to meet.

2 The concept of the nucleus There has always been some debate in intonological circles as to the existence or otherwise of a nucleus. However, it has never been denied that in many languages, the final accented syllable and unstressed adjacent syllables have some special form, and often express some special function. Although Palmer (1922) is the originator of the term "nucleus" which is the "stressed syllable of the most prominent word in the Tone-Group", Cruttenden (1990) points out that Alexander Melville Bell and David Charles Bell, (the first of whom taught Henry Sweet) had used the term "emphasis" with a meaning very close to the term "nucleus" as it is used within the British school today: "Thus, as it were in a picture, the more essentia] parts of a sentence, are raised, as it were, from the level of speaking; and the less necessary, are, by this means, sunk into comparative obscurity." [punctuation as in original] (A. Bell, 1835:xlii-xliii, quoted by Cruttenden, 1990:3).

3 Since it is specifically claimed that it is the intonation contour on and around the "nucleus" which differentiates polar questions from statements, the focus in the following chapters will be on that part of the intonation contour.

3 Chapter outline The British-school definition of the "nucleus" is discussed in chapter 2 with respect to its domain. It is examined within a componential analysis of the Tone-Group, proposed by Palmer (1922). Such an analysis of the Tone Group (or Tone Unit as it is often referred to), although with a slightly different inventory of component parts, continues within the school today; see Crystal (1969), O'Connor and Arnold (1973), Gimson (1980), Couper-Kuhlen (1986) and Cruttenden (1986). However, an area of little consensus is the mapping of the nuclear tone onto the nuclear syllable and tail. In order to reach a maximum intersection across the studies investigated, a single indivisible component in place of the latter two is therefore proposed as the domain of the nuclear tone. This structural analysis is adopted in chapter 3, where a first attempt is made at describing Palermo Italian intonation. Two questions are addressed: (i) what intonational form is used to signal interrogation? and (ii) can this form be adequately described using the model proposed? A partial answer is given to (i) and a negative answer to (ii), which leads to the investigation in chapter 4 of whether a more appropriate model has been adopted by Italian auditory analysts. In the selection of Italian studies surveyed, the intonational form signalling interrogativity is shown to differ from one variety to another and, in all but one case, from Palermo Italian too. All of the studies appear to rely on models which are rooted in the tradition of analysis in other languages, predominantly English or French. One problem with the British analysis is shown to be the fact that each component is characterised by a combination of perceived pitch and stress, there being no strict separation between the pitch contour and the rhythmic structure arising from the choice of words. An alternative approach to the componentialisation of intonation contours, the theory of autosegmental phonology, is discussed in chapter 5. Goldsmith (1976) lays out the principles of this theory, developing his analysis with examples from "tone languages", and tentatively applying the approach to English as an "intonation language". One of the main tenets of the theory is a formal separation between tune and text; tones in the tone tier and vowels in the phoneme tier are synchronised at strategic points by means of the principle of association. This has instigated a considerable body of subsequent work in which attention is directed not only to the underlying phonological association but, more specifically, to the alignment of the segmental structure with peaks in fundamental frequency (and, to a lesser extent in perceived pitch). In this chapter, work by Hirst, Ladd, Bruce and Garding and Pierrehumbert and Beckman is surveyed with the intention of distilling the aspects of each study which might prove useful in accounting for the Palermo data.

4 Chapter 6 returns to the domain of the "nucleus", first discussed in chapter 2. This time, however, the assumption that the nuclear tone is mapped onto the nuclear syllable and what follows it (the tail) is brought into question. It is shown that, even within the British school, the left edge of the nuclear domain is not as fixed as is often suggested. Kingdon (1958) observed that a contrastive prehead, which enhances the already emphatic nature of certain nuclear tones, can be realised in the absence of a prehead syllable. The preaccentual initial pitch excursion in nucleus-initial tone units cannot, in fact, be treated as constrastive by the majority of analysts within the British school; such pitch movement is considered to be an onglide and has no phonological status. This is also shown to be true for a number of autosegmental pitch accent analysts. Ladd, for instance, is in accord with the British view, allowing only for trailing tones in bitonal pitch accents (i.e. the first tone in the pitch accent is always associated with the stressed syllable). However, certain analysts within this tradition (such as Pierrehumbert and colleagues) can accommodate this initial pitch as a leading tone of a bitonal pitch accent, viz. L+H* and H+L*. Whereas in the particular context he discusses, Kingdon considers the contrastive pitch to be prenuclear, delaying the onset of the nuclear tone, Pierrehumbert (1980) incorporates the leading tone into the domain of the final pitch accent (which involves what British analysts call the nuclear domain).The fact that the functional load borne by leading tones is smaller than that borne by trailing tones is given as the reason why many analysts ignore contrastive pitch in this early position. A solution is proposed for English Pitch Accents, allowing the nuclear domain to be extended to the left, but treating this leftwards extension as a proclitic element. This is achieved by means of an enriched Pitch Accent structure consisting of two levels: Supertone and Tone. A leading Tone in a Pitch Accent is dominated by a weak Supertone, whereas starred and trailing Tones are always dominated by a strong, branching left-headed Supertone. The strong Supertone is considered to be the core of the Pitch Accent. In Chapter 7, a corpus is discussed whose aim was to record a number of statementquestion pairs which differ only as a function of their intonation contour; such recordings were carried out in order to maximise naturalness. A number of utterances are singled out for detailed description, especially with regard to the synchronisation of peaks and troughs in the fundamental frequency traces with segmental landmarks in the utterance. In certain cases, such peaks are taken to correspond to H and L tones within the autosegmental pitch accent approaches referred to in chapters 5 and 6. Such an approach is shown to account for the data presented, as well as for phenomena noted in chapter 3. A model is developed within the autosegmental framework which captures both "contextually determined" variation and distinctive contrasts within the language. Such a framework makes full use of leading as well as of trailing tones. The Pitch Accent structure proposed for Palermo Italian reflects the high functional load borne by leading tones; they are shown to be better accounted for as part of the core of the Pitch Accent (the part dominated by the strong supertone) rather than as proclitic elements. Thus, in Palermo Italian, the Pitch Accent node does not branch; it dominates one Supertone

5 node which, in turn, dominates maximally two Tones. The Supertone may be right or left headed, and Pitch Accents with leading tones are represented with a right-headed Supertone. Furthermore, differences in detail between the alignment of tones in Palermo Italian and English are accounted for by the differing Pitch Accent structures proposed in chapters 6 and 7. Chapter 8 sums up the findings of the previous chapters and points to the wider implications of the research undertaken.

II. The British nucleus-plus-head approach

1 Introduction The purpose of this chapter is to describe the approach to intonation adopted by the British school, and to point out its advantages and disadvantages for the description of a corpus in a language which has hitherto not been analysed. The British approach to intonation, which has been used extensively for the transcription not only of English but of a number of other languages, relies on a long tradition of auditory analysis. It is taken as a starting point here because auditory perception is a useful tool for finding linguistically relevant categories. The term 'nucleus-plus-head' refers to the development within that approach in which tone units are analysed into component parts ('functional units') rather than indivisible tunes. Such componentialisation began with the work of Palmer (1922) who subdivided intonation contours into a prenuclear portion which he called the 'head', a 'nucleus' and a 'tail'. Further subdivisions of the prenuclear portion of the contour into 'prehead' and 'body' were introduced by Kingdon (1958); O'Connor and Arnold developed the theory for didactic purposes, reconstituting the components into a number of typically occurring 'tunes', whilst acknowledging the importance of the constituent analysis as a starting point. Gimson's (1980) introductory text on English pronunciation also takes a didactic line in his chapter on intonation, although he does not do any reconstitution. For the description of fine phonetic detail of the kind found in spontaneous corpora, the whole theory was elaborated by Crystal (1969). The most recent comprehensive work, that of Couper-Kuhlen (1986), continues along similar lines to Crystal, describing recorded conversations, but to a greater degree of precision. Another recent study by Cruttenden (1986) describes and develops the British approach in the context of a number of other theories. Common to all but the very early work within the British school outlined above, is the definition of the tone unit. This will be our starting point. We shall then take pairs of constituents in turn and examine the evidence for a clear-cut componential analysis, both in terms of whether the distinction between them is unambiguous, and whether it is appropriate to disregard any transitional phenomenon which, by definition, occurs outside the domain of both of the adjacent constituents. Finally, we shall discuss which aspects of the theory are retained, and which are modified, for the analysis of Palermo Italian reported on in the following chapter.

8

2 The tone unit According to Crystal the tone unit is a stretch of utterance consisting of an obligatory element, the nucleus, and three optional elements, the prehead, the head and the tail. He represents the structure of the 'tone unit' as (Prehead) (Head) Nucleus (Tail) where the optional elements are in brackets. The definitions of the four components of the tone unit rely on a set of underlying assumptions concerning the rhythmic structure of English, in particular, a distinction between unstressed, stressed and accented syllables. Crystal makes this distinction explicit. A 'stressed' syllable is a syllable which is perceived as prominent in relation to the other syllables in a given tone unit. This prominence is due to any or all of a number of phonetic features including increased loudness (acoustic correlate: amplitude), increased length (acoustic correlate: duration) and unreduced vowel quality (acoustic correlate: spectral profile), and may involve pitch prominence (acoustic correlate: FO). Crystal (1969:162) points out that we probably " 'read in' rhythmic regularity" when we hear an utterance, i.e. the fact that we expect a relatively regular rhythm leads us to perceive regular peaks of prominence which may have no acoustic correlate at all. He warns the reader not to be sidetracked into a detailed discussion of rhythmic prominence. This is probably because his approach is auditory, and an investigation into the complex interactions between acoustic correlates of stress would be well beyond the scope of such a study1. Crystal also distinguishes syllables which are simply stressed from those which have the additional specification of being accented. An accented syllable is considered more prominent than a plain stressed syllable by virtue of the fact that it involves a pitch obtrusion either a jump or glide. The extra acoustic correlate of accent is thus FO change. Unstressed syllables may be pitch prominent, but these are not considered accented. This will be discussed further in section 3.1. A tone unit with the full complement of Prehead (P), Head (H), Nucleus (N) and Tail (T), as in the example below, can be used to illustrate the domain of each component2. The 'nuclear syllable'3 is generally said to be the last accented syllable in a tone unit from which or upon which there is a pitch change; it is usually, although it need not necessarily be, a lexically stressed syllable, (i.e. a syllable marked in the lexicon as potential bearer of promi-

1

It is also a reflection of the fact that the rhythmic structure of English, stress-timed with the head of the foot on the leftmost syllable, is taken for granted. 2 The two parallel lines represent the top and bottom of the speakers pitch range: the heavy dots a 'stressed' syllable and the small dots an 'unstressed' syllable. Accented syllables are not specially marked, because pitch obtrusion is determined relative to preceding or following stressed and/or unstressed syllables. 3 The fact that we use the term 'nuclear syllable' rather than 'nucleus' will be explained later.

9 nence). In the above example, this is the first syllable of 'reason1, on which there is a pitch glide. She

~

(P)

COULDn'thave FOUND a BETter REAson

*

'

*

(H)

•

.

•

"

(N) (T)

The nuclear syllable can be distinguished from plain accented syllables in a number of ways. Syntagmatically, it is the last accented syllable of the tone unit; paradigmatically, the nuclear syllable can carry a greater range of pitch movements (if no tail is present) than other types of accented syllable. However, it is more than just the last accented syllable: the nuclear tone, which it (partly) carries constitutes the 'peak of prominence' of the tone unit (Crystal: 1969:209). This concept underpins the British work described here; the primacy of the nucleus (although involving the nuclear syllable and the tail) has led to the categorisation of tone units into major groupings according to nucleus type. The tail is made up of all syllables following the nuclear syllable, which, in the above example, is only one, the last syllable of 'reason'. The head begins on what Crystal refers to as the 'onset', the first accented syllable of the tone unit. In the above case the head begins on the syllable 'could' and extends up to but does not include the nuclear syllable; the end of the head is on the last syllable of 'better'. The prehead comprises all the syllables before the onset; here there is only one syllable in the prehead, the syllable 'she'. It is possible for a stressed syllable to occur in the prehead provided that it is not accented. It is also possible for a prehead syllable to be pitch prominent, provided it is not stressed. A prehead may occur immediately before a nuclear syllable with no intervening head. In the next section, we shall discuss the merits and demerits of this four-part componentialisation of the tone unit.

3

Componentiality

As we have seen, the definition of each component relies on prominence lent by a combination of rhythm and pitch. In the following section, we shall take each adjacent pair of tone unit constituents and examine whether they can be consistently distinguished and whether the boundary between them is clear cut.

10

3.1 The prehead-head distinction The distinction between prehead and head is particularly problematic in cases where the prehead has a different pitch height from what follows. The inherent difficulty involved in deciding whether a syllable is stressed or not, is increased if the pitch of the syllable obtrudes; pitch obtrusion lends prominence and can lead to the perception of stress. There is one case where this cannot happen: stress is precluded if the vowel quality of the syllable nucleus is reduced. However, since not all vowels undergo the same degree of reduction, there are often inconsistencies in transcription. An example of such inconsistency is shown in the following two examples: (i)

IN

LANguage

(ii)

T h e LANguage

Although the configurations are similar in shape, they may have a different transcription simply because of the spectral quality of the vowel in the first syllable. In the first syllable of (i), 'in', the vowel quality is appropriate for either a stressed or unstressed syllable; the prominence lent to the syllable by the pitch obtrusion might therefore lead to the transcription of an accented syllable in this position, thus making the first syllable the onset of a head 4 . This cannot happen in (ii) because the reduced vowel quality in 'the' (with a schwa) precludes the analysis of the high pitched syllable as stressed. The pitch prominence cannot therefore be interpreted as an accent; the first syllable can only be a prehead. Setting aside the problems in distinguishing stressed, unstressed and accented syllables, another question to be asked of a componential analysis is whether it has the facility for elements of a tune to be carried by different components. For instance, we take a look here at the interchangeability of the prehead and the head. O'Connor and Arnold's (1973) tune, the 'Low Bounce' involves Prehead

Head high

high

Nucleus low rise low rise

The prehead may be low only if followed by a high head: low 4

high

low rise

This example is taken from the Esprit SAM project "numbers passage" on CD Rom (Eurom-0); transcriptions of the same recording by a number of analysts represented the intonation of "In" as either a high prehead or a high head. One analyst transcribed a fall-rise nuclear tone starting on "In". There was no consensus transcription.

11 This tune does not occur without a prenuclear component. For O'Connor and Arnold, the presence of high pitch just prior to the nucleus is necessary to functionally demarcate this tune from another, the 'Take Off, where there is no such high pitch: "...no criticism is implied, such as is found with the Take-Off..."(1973:62) and "when there is no head, the High prehead is used to avoid the scepticism of the Take-Off." (1973:65). Thus, if four tone units are compared:

(i) (ii) (iii) (iv)

Prehead high low high low

Head

high high

Nucleus low rise low rise low rise low rise

(i) would be considered functionally closer to (iii) and (iv) than to (ii). Even though prehead are identical in (ii) and (iv), (iv) is grouped with (i) which only has same nucleus. It is high pitch before nucleus which is important. This fact is not captured in a purely componential account, even though, intuitively, it appears to be the falling-rising combination which is important. In other words, whereas a holistic tune approach might describe a 'falling-rising' tune, a componential approach would proffer different analyses, depending on whether the rising movement were carried by stressed and/or unstressed syllables. The advantage gained by applying the British nucleus-plus-head approach is that, by dividing tone units into component parts, all tone units with a particular nuclear tone can be grouped together. It is argued (inter alia, by Crystal: 1969) that this is the most important generalisation to be captured. In other words, in (v)

-

-

low rise

there is only a rising movement, but this belongs to the same major class of tone unit as (i) through (iv)5. A holistic tune approach would not capture this. In order to investigate the role of the nuclear tone in grouping tone units, a closer look at the part of the tone unit which carries the nuclear tone is in order. This is done in the following section.

3.2 The nucleus-tail distinction In this section, we discuss the definitions offered by various intonologists of the nucleus, nuclear syllable and tail. We examine the distinction between the nuclear tone as a pitch movement and the domain over which it is manifested. Finally, we discuss the advantages

What happens just before the nucleus will be discussed in detail in chapter 6.

12 and disadvantages of a very narrow phonetic transcription method for the characterisation of nuclear tails.

3.2.1 The nucleus, the nuclear syllable and the nuclear tone The literature on intonation makes use of the term 'nucleus' to denote either the nuclear tone, the tonal (or pitch) movement, or the nuclear syllable, described above as the syllable in the tone unit on which or from which there is a pitch change. Most analysts within the British tradition claim that the nucleus occurs on the word which is focussed, most important, new, etc. However, attention will be concen-trated here on phonetic aspects of the nucleus rather than on the reasons a speaker might have for placing the nucleus in a particular position in a tone group. O'Connor and Arnold (1973) and Cruttenden (1986) consistently use the term 'nucleus' for the syllable and 'nuclear tone' for the tonal movement. However, Crystal(1969), Gimson (1980) and Couper-Kuhlen (1986) all use the term 'nucleus' to mean either. The following table shows how 'nucleus' and other terms are used:

6

AUTHOR

SYLLABLE

TQNAL MOVEMENT

O'Connor and Arnold (1973)

nucleus (pl4)

nuclear tone (pl5)

Crystal (1969)

nuclear syllable (pl42), nucleus (p208) 7

nucleus (pl42) 6 , nuclear tone (pl42)

Gimson (1980)

nuclear syllable (p267), nucleus (p267) 9

nucleus (p265) 8

Couper-Kuhlen (1986)

nucleus (p79) 10

nucleus (p86) u , nuclear movement (p86)

Cruttenden (1986)

nucleus (p56) 12

nuclear tone (p57)

Crystal (pl42): 'This obligatory, and usually kinetic tone I shall refer to as the nucleus of the tone unit (or nuclear tone)' 7 Crystal (p208): 'Prehead Head Nucleus Tail' ® Gimson (p265): '... with the nucleus (falling, rising or a combination of the two) on the appropriate syllable' 9 Gimson (p267): "When syllables follow the nucleus - the tail...' Couper-Kuhlen (p79): '...the nucleus is the most prominent syllable in a tone unit' Couper-Kuhlen (p86): '...nuclei - unless they are monosyllabic - are more often than not 'spread' over the tail or part of it' 10 Cruttenden (p56): '...The occurrence of similar tones starting from the nucleus'

13 In the remaining analysis, the term 'nucleus' will be avoided in favour of the less ambiguous, even if also less eloquent, 'nuclear tone' and 'nuclear syllable'. Having established the nuclear tone as a pitch movement or configuration, we now examine its domain.

3.2.2 The domain of the nuclear tone Cruttenden (1986:57) uses the term 'nuclear tone' to refer to "what starts at and follows the nucleus" (emphasis added). In other words, the nuclear tone occurs on both the nucleus and the tail. According to Crystal (1969:223), "by far the most frequent pattern" of the tail is when it continues the direction of the nuclear tone, although it may also begin by doing so and then level out. He adds that levelling of the tail is uncommon in rising tones. This levelling occurs either because the speaker's voice range limits have been reached, or because the speaker chooses to level out the tail in order to convey a particular attitude (such as irony, sarcasm or boredom 13 ). Apart from cases where levelling occurs, he claims that tails are 'usually non-distinctive'. Gimson appears to differ from Crystal in two respects: (i) He claims that 'unaccented syllables following a falling nucleus (= nuclear syllable) remain on a low level', they do not therefore follow in the same direction as the nuclear syllable. (ii) In the case of rising nuclei he claims that the unaccented syllables not only continue but may also 'effect' (1980:267) the rise, (i.e. there is not necessarily a rise on the nuclear syllable itself). In fact, he claims that, if there is a tail, the pitch movement rarely begins on the nuclear syllable: 'with a tail, the rise is achieved by means of a relatively low pitch on the nuclear syllable with an ascending scale on the following syllables' 14 . The situation is further complicated since he also adds that when a low rising glide occurs on a short syllable of the type described above and there is no tail [see la], it may be substituted by 'a relatively high level pitch in relation to a preceding low pitch [seelb], or even a slightly lowered level pitch in relation to a preceding mid or high pitch' [see lc] (1980: 267):

13 He is referring to a phenomenon which has since been widely referred to as stylised intonation, as discussed in detail by Ladd (1978). 14 In such a case, if the head also happens to be at the same pitch level as the nuclear syllable, the latter would be no more pitch prominent than the first tail syllable. We can only assume that, in such situations, in such situations, the nuclear syllable is recognised as such either because it is the lexically stressed syllable (assuming native speakers know which syllable in a word is lexically stressed), or if it is not, e.g. in 'Did you say intralinear?' where the nuclear syllable is 'tra', because it is perceived as prominent in some way other than by pitch (e.g. by loudness or vowel quality).

14 'CAN she

COOK?' «

la

•

lb

lc

la, lb and lc are all transcribed intralinearly with a low rising nucleus on the syllable 'cook'. Similarly, he transcribes both a low falling glide (see 2a) and a 'low level pitch in relation to a preceding higher pitch'(see 2b) (1980:267) as a low falling nucleus in his interlinear transcription: 'WHAT have you GOT?' •

.

^ 2a

.

.

i

2b

On the grounds of functional similarity, he implies that patterns la, lb and lc are variations on one pattern and 2a and 2b are variations on another. However, the interlinear representations of lc and 2b are identical. There is an implication here that the starting point of the nuclear syllable, when short, can carry pitch movement which would have been attributed to an onglide. In short, lc has a fairly low pitch which, it appears, substitutes for the jump down to low for the start of the low rise as well as the rise up from the low start of the nuclear tone. Similarly, 2b has a fairly low pitch which substitutes both for a low start (relative to the previous syllable) and a low end, of a low fall. This shows that the boundary between nuclear and prenuclear pitch phenomena can be somewhat blurred. Couper-Kuhlen (1980) recognises that too little importance has been attributed to the tail, since it actually 'contributes to the shape and identification of the nucleus' (=nuclear tone), but agrees with Crystal stating that 'rises are typified by upward pitch movement on or beginning on a prominent syllable' (= the nuclear syllable). Some of the differences in where the rise or the fall exactly occur may depend on the segmental constitution of the nuclear syllable. According to Crystal, a pitch jump is functionally identical to a glide and occurs when the voiced part of the nuclear syllable is too short to carry the pitch glide (e.g. a syllable containing a short vowel followed by a voiceless consonant). Gimson (1980:267) gives both possibilities, a glide or a pitch jump, for the following tone unit: It was YESterday

or

It

was YESterday

15 making no mention of a functional difference between the two. If Crystal's assumption that the occurrence of a glide is phonetically determined is correct, then we must assume the overall rate of delivery of the tone unit, the actual duration of the short syllable or the duration of voicing to be possible determining factors. In the following example, Gimson shows a glide on the nuclear syllable 'rain', leaving little room for pitch movement on the tail syllable 'ing'15. It

.

was RAINing

.

V

Whereas in the following, he shows no glide at all on the nuclear syllable. This is not surprising, considering the segments involved in the nuclear syllable: a voiceless syllable onset, a short vowel and possibly a devoiced nasal offset: ARE you

COMfortable?

Cruttenden argues that although the occurrence of a pitch jump or glide normally depends on the phonetic criteria outlined by Crystal (e.g. duration of voiced stretch), such realisations are not obligatory, and a speaker can choose to use a jump from a long vowel to a following syllable and a glide on a short vowel (1986:54) (see also Bolinger 1985:29). Jumps which occur where glides are expected sound 'abrupt', whereas glides occurring instead of jumps tend to sound either 'soothing' or 'reproachful'. He adds that whereas glides are frequent in English, jumps are predominant in German. We shall return to the issue of glides and jumps, examining them from Couper-Kuhlen's point of view, in section 3.2.3 below. In the above examples it has been shown that, for whatever reasons, the nuclear pitch change can take place on both the nuclear syllable and the tail. If it takes place on neither it is because both are uttered on level pitches, the change in pitch being a jump from the level of the nuclear syllable to the level of the tail. O'Connor and Arnold (1973:19) provide such an example: DON't be SILly •

15

•

He actually marks no movement at all on the tail syllable.

16 There has been little discussion so far of cases where such an important phenomenon in the utterance (the pitch change) takes place in a position which can only be analysed as being between two elements of the tone unit., in this case, between the nuclear syllable and the tail. This is because each component of the tone unit is seen to be independent, and no room is made for transitional phenomena; they are, in fact, ignored.

3.2.3 Couper-Kuhlen: Steep slope, gradual slope and flattening The most detailed auditory approach develops further the notation of Crystal. In a section on tails, Couper-Kuhlen (1986:86-88) discusses steep slope, gradual slope and flattening. These will be examined in turn.

3.2.3.1 Steep slope She discusses the 'slope' of the 'prolongation of the nuclear movement' (1986:86), which, if we also take into account that the section is entitled 'tails', we can assume to mean 'the movement on the tail'. However, steep slope occurs when the pitch falls steeply on the nuclear syllable or from the nuclear syllable to the first syllable of the tail, by which time the entire pitch movement has been accomplished. Since there is no pitch change on the tail, (in fact, all the examples given have a low level tail), it might be preferable to apply the adjective steep not to the pitch movement on the tail but rather to that before the tail. Again, the transition between the nuclear syllable and the tail is at issue. In the following example (p87) of steep slope,

CARE

for

that

child

the pitch movement is accomplished by a glide on the nuclear syllable plus a jump down from the end of the glide to the first syllable of the tail (within the encircled part above). Couper-Kuhlen (1986:87) marks steep slope but does not mark the width of the nuclear movement. In her diacritic tonetic transcription system she is able to mark nuclear tones as wide [ w ] and narrow [ n ] 1 6 . If neither of the diacritics are present, which is the case here, then we assume the pitch movement to be of 'normal width'. However, earlier on in the chapter (pp79-80) she adheres to the view that the tunes of miner [ \ ] and cutter [ .] are intonational 'allotones'. If this is so then the term 'wide' might be better applied to the whole portion enclosed in the circle, otherwise 16

Couper-Kuhlen borrows these diacritics from Crystal (1969). However, whereas Crystal uses them to refer

to stretches of utterance, Couper-Kuhlen uses them to refer to nuclear tones.

©o Miner

and

17

CUTter

must be classified in different ways, the first as having a 'wide nucleus' and the second as having a 'steep slope on the tail'. This is counter-intuitive, given the extremely different segmental make-up of the examples. It is difficult to place the emphasis on speaker choice in the face of such segmental differences; the pitch jump on 'cut' in 'cutter' must largely be due to the short voiced stretch. Even if we take into account Cruttenden's observations about jumps sounding abrupt and glides sounding soothing or reproachful, the fact remains that the segmental content of the two words is not at all similar and this undoubtedly has some effect on the realisation of the nuclear tone. It is perhaps the case that a glide-jump distinction is only available in less extreme cases, such as the example discussed by Gimson of "It was yesterday". However, even there, a glide might be produced as a function of rate of delivery, another interacting factor which would increase the duration of the voiced stretch and thus favour a glide. Crystal (1969:151) marks the whole portion of the tone group reproduced here as wide. NOT sucCEEDed

But, if we count 'wide' to cover the encircled portion, then 'steep slope' would be a redundant feature in the classification of nuclear tones since a wide high falling nuclear movement is identical to a high falling nuclear movement with steep slope. Given the great effect of segmentals on the wide-steep distinction, one term covering both would make fewer non-functional distinctions. This approach is preferred in the study reported on in chapter 3.

3.2.3.2 Gradual slope Couper-Kuhlen's note: 'Obviously length of tail interacts with slope. A very steep descent (or rise) on a lengthy tail will often mean that the final syllables level out' (1986:86-87) does not seem to be consistent with the fact that, in most of her examples, the fall is completed by the first syllable of the tail, the length of the tail being therefore irrelevant. She refers to gradual slope when the tail continues the pitch movement. It appears that, in her view, pitch movement only occurs on the tail in the case of gradual slope. She also claims that gradual slope is 'the unmarked variant, or the norm for tails' (p87). Most British intonologists agree that this is the case for rising nuclear tones but they would not all argue that the unmarked

18 falling nucleus is spread over the tail, at least not to the same degree as rises. O'Connor and Arnold (1973:15) give the example of SEVenty

where the fall is completed within the circle as in the above examples. It could be argued that such a sharp fall is typical of their clipped peremptory style of speech, and that CouperKuhlen is describing a less 'advanced RP' variety. The sample of speech provided in the accompanying tape bears this out. However, in her section on simple falls (p90), she herself gives only examples where the pitch movement is completed by the first syllable of the tail. Gradual falls and rises include, by definition, the syllables in the tail, and may therefore be levelled out if the tail is long.

3.2.3.3 Flattening What Crystal refers to as 'levelling' (1969:223-224), discussed above, corresponds to a certain degree to Couper-Kuhlen's 'flattening'17. However, it seems that flattening may only occur where there would otherwise be gradual slope. In the example below Couper-Kuhlen (1986:88) claims that the final syllable would have been lower if the tail had not been marked as 'flattened'; this lower position is indicated by the arrow:

Which WERE FALIible / N

unmarked level of syllable

In the same section on tails she also mentions 'endpoint', 'the pitch level which the endpoint of the tail reaches' (p87). She argues for the marking of extremely high or low endpoint. She argues that even in cases of wide pitch movement the endpoint may 'stop short of the extreme' (p90). However, a matter of pages earlier, the following example is marked with 'extremely low endpoint' (p87):

17

However, if levelling refers mainly to stylised contours, such as those discussed by Ladd (1978), flattening can refer to a type of contour which has been described by Gussenhoven (1983) as 'half completion' and by others as 'curtailed' contours. See Johnson and Grice (1990) for a discussion of the types of contour involving level or near-level tones.

19

supPORT of ERin

•

V

.

.

NOW DAVid

.

.

.

whereas the following is not (p87) 18 : CARE for that child

although the end of the tail seems to be at the same level in both examples 19 . Although this section has been concerned with Couper-Kuhlen's description of tails, it has been shown that the nuclear syllable and tail function as a more or less indivisible unit. The shape of either can be greatly influenced by the segmental structure of the syllables concerned, and the shape of each is dependent to a large degree on the shape of the other. Such problems lead to the proposal in section 4 to treat them as one indivisible unit rather than two.

3.2.4 The interdependence of nuclear syllable and tail There is a difference in status between the nuclear syllable and the tail in that the former is the syllable which is psychologically most important (i.e. the lexically stressed syllable of the focussed word). The nuclear syllable is obligatory, whereas the tail is not. This is, of course, not to say that the tail doesn't carry distinctive pitch movements: complex tones which have been somewhat neglected in this account almost invariably involve the tail (if there is one). The importance of the nuclear syllable and the relative lack of importance accorded to tails may have been behind Crystal's statement that the pitch contours of tails are 'automatically determined by the direction of the nuclear tone' (1969:223). However, the analysis is simplified if we see the nuclear tone is an abstract pitch movement which must be synchronized with a segmental layer made up of both the nuclear syllable and the tail. In this case, the pitch of both the nuclear syllable and the tail is determined by the direction of the nuclear tone. We shall return to this issue in section 4.

18 Couper-Kuhlen uses the diacritic [I] to indicate steep slope and [l]to indicate low endpoint, which, in the case of falls, are placed under the diacritic for the falling tone [" ]. 19 It is always possible that different pitch ranges are indicated by the upper and lower lines, although this is not explicitly stated. If this were the case, perhaps a global range parameter might be a more meaningful way of expressing the distinction between the two examples.

20

3.3 The head-nuclear syllable transition When classifying heads, Crystal (1969:229-231) takes into account the pitch level of the onset. He also uses the pitch at the beginning of the nuclear tone in order to create subcategories of heads requiring 'marked' pitch height on the nuclear syllable. However, this does not affect his major categories; even if the pitch of a nucleus following a rising sequence of stressed syllables is lower than the preceding stressed syllable, the head is still categorised as rising rather than rising-falling. Couper Kuhlen (1986:84-86) does not consider the pitch of the beginning of the nuclear tone as significant, hence her comment that monosyllabic heads cannot be classified as rising or falling 20 ; a step up from the onset syllable in a monosyllabic head to the nucleus would not make the head a rising one. She states clearly that her head classifications are 'wholly independent' of nuclear type'. By contrast, the issue of whether the nuclear tone classifications are wholly independent of head type is less clear-cut. In the following example transcription which she analyses into two tone units (1986:80), the second one beginning on which: and THIS is SOMEthing which I think

In the first tone unit where this is the nuclear syllable, she states that the rising pitch movement from and to this is irrelevant in determining whether the nuclear tone is classed as falling or rising. It is, in fact, described as a falling tone because of the following low pitch on is. Likewise, the nuclear tone in the following tone group is determined by the pitch movement from 7, the nuclear syllable, to think, the tail. However, Couper-Kuhlen's system of relative pitch height, like Crystal's, takes into account the transition between the head (or prehead) and the nuclear syllable: "nuclear pitch height is relative to the height of the preceding syllable" (1986:91). This is also true of Cruttenden. Thus, the boundary between the head and the nucleus is seen to be irrelevant in one regard and important in another. However, it is never considered to be as important as the direction of pitch movement on and after the nuclear syllable. It is this point which must be emphasised, and which Couper-Kuhlen was most probably indicating in stating that "What the pitch of the voice does on the syllable before the prominent syllable is of no relevance in determining whether the nucleus is falling or rising." (1986:80). It is of no relevance in signalling nuclear tone direction but of some importance in signalling nuclear tone height. The hierarchy of importance relies on the fact that direction is more important than height of starting point. This could be said to be the consensus view amongst all nucleus-plus-head approaches.

20

unless, presumably, they actually carry a pitch glide

21

4 A working model For the purposes of the Italian study described in the next chapter, the principle of componentiality is strictly adhered to; this is done in order to allow for an investigation of the different components in isolation, as well as the transitions between them. Owing to the problems encountered when treating the nuclear syllable and tail as separate entities (reported on in 3.2), they are collapsed into a 'nuclear unit' which corresponds to the single indivisible domain of the nuclear tone; it encompasses the nuclear syllable, the tail and the transition between them. Such a proposal reduces the tone unit to three constituents: (Prehead) (Head) Nuclear Unit The mapping between the segmental structure and the tune: the prehead tune, the head tune and the nuclear tone, is thus simplified, and a componential analysis can more easily be performed. This type of division is not new, however. Within a different, although British, tradition, Halliday (1967) 21 proposes an indivisible unit, 'the tonic' as the domain of the nuclear tone; in his terminology the 'tonic' is the domain of the 'tone'. He does not subdivide the 'tonic' into independent components. The 'tonic syllable' is equivalent to the nuclear syllable, but it is still part of the 'tonic', there being no posttonic element or tail. The nuclear unit proposed here is thus equivalent to Halliday's tonic 22 . In order to elucidate the nature of the transitions between the head and nuclear syllable (discussed in 3.3), a syntagmatically non-relational method of classifying the height of nuclear falls and rises is adopted. For instance, in this method, 'high' would refer to the position within the speaker's range 23 . In the proposed system, the height refers to the starting point of the nuclear tonal movement; thus [. *\] is a high fall, as is [,*\], and [,*\] and [. *\] are both mid (not low - as they begin on a mid pitch) falls. Couper-Kuhlen's fine distinctions regarding width and endpoint (discussed in 3.2.3) are collapsed into a single dimension of width which refers to the whole of the nuclear unit. The endpoint is not recorded separately, but rather inferred from a combination of starting point height and width. Heads and preheads are defined as ending on the syllable before the nuclear syllable and their end height as well as starting height is recorded. The distinction between head and pre21 Halliday's work pays more attention to functionality than to minute phonetic detail. The phonetic detail he does uncover is often neglected, together with his ideas on the relationship between intonation and grammar, which have suffered much criticism from Crystal, especially at the time of writing his 'Prosodic systems and the Intonation of English'. Nontheless, Halliday's tonic has been retained by Brazil (1985) and, conceptually adopted by Nolan (1984) in the form of a "nuclear accent unit". 22 Similar indivisible nuclear units were used in American studies: Pike (45) and Hockett (58). See Ladd (1980) for a discussion of these. 23 This is an adaptation of O'Connor and Arnold's method of describing nuclear pitch height, to be discussed further in chapter 3

22 head has been shown to be difficult to maintain in cases where the pitch of a syllable obtrudes in relation to what precedes or follows it (see 3.1). However, an attempt will be made to maintain this distinction, whilst bearing in mind that the division between the two categories is by no means watertight. The next chapter will apply this working model to a corpus of spontaneous speech in Palermo Italian, and will discuss in particular the nature of the transitions between the different components which have been recorded and what functional value they have.

III. A British-style analysis of Palermo Italian

In this chapter, reference is made to the author's auditory analysis of a spontaneous corpus of Palermo Italian. This analysis was originally carried out in partial fulfilment of the requirements for the degree of MA at Reading University. The transcriptions were performed within the British auditory tradition, assuming that, with certain adjustments, Palermo Italian intonation could be analysed using the techniques and principles developed within this approach. There were no considerations of pitch accent or autosegmental analysis. It is for this reason that the results are referred to here, since they will be compared to an instrumentally corroborated autosegmental pitch accent analysis of Palermo Italian in chapter 7. Since the transcriptions were performed as part of another degree program, they will only be summarised here. They will serve as a catalogue of spontaneous intonation patterns. In the interests of objectivity, the interlinear (•.) transcriptions made for the purposes of the MA dissertation have not been altered (a) with regard to the position of a dot in the vertical interlinear space and (b) the decision to label a syllable stressed or unstressed (one exception to this is pointed out in the text). Although the basic analysis remains unaltered, some of the data are presented in a different light; a number of previously unnoticed inconsistencies are pinpointed for further discussion in chapter 7.

1

Introduction

The object of study is the intonation of the local variety1 of Italian spoken in Palermo, the capital of Sicily. Of particular interest is how speakers of this variety signal interrogation in yes-no questions. As in most varieties of Italian, yes-no questions and statements are distinguished solely by their intonation contours, there being no syntactic or morphological markers for interrogation in this context. Two basic questions arise: What intonational form is used to signal interrogation? Can this form be adequately described using a model of intonation based on the nuclear tone? The first of these questions is given consideration in section 3 below. It is then given a more precise answer in section 4. The second is addressed is sections 4 and 5. 1 The concept 'local Italian' is invoked here because the Italian spoken in Palermo differs substantially from the Italian spoken in other parts of the region (Sicily), both in terms of its segmental phonological characteristics and its intonation. Thus, the concept of "regional Italian" as discussed, for instance, by Lepschy and Lepschy (1977) is insufficiently precise to describe the variety under investigation.

24

1.1 Corpus details A problem generally encountered when conducting a corpus-based analysis of interrogative utterances relates to their low frequency of occurrence, making a sufficient number for an adequate count very difficult to obtain. Fries (1964) overcame the problem of recording yesno questions by recording a number of television and radio programmes in which a number of panellists attempt to discover the precise vocation, occupation or special activity of another person who may give no information other than yes or no answers 2 . The corpus thus contained a high proportion of yes-no questions. Lee (1980) has since discussed Fries' results, suggesting that the type of intonation used might depend, amonst other things, on the distribution of yes-no questions thoughout the data. In most English yes-no questions, however, morphological and syntactic markers operate in conjunction with intonation, although where the context clearly requires a question, even a syntactic statement with falling intonation may be interpreted as a question. In contrast to English, Palermo Italian makes exclusive use of intonation as a marker of interrogation. Nontheless, it is acknowledged here that interrogation may not always be signalled explicitly, expecially where contextual cues are given3. Since the recording of a radio or television programme in Palermo Italian was not possible, recordings were made of a similar type of game played by a number of Palermo Italian speakers in the home of one of the pertecipants. A number of recordings were made of different occasions during the Christmas vacation, a period during which a number of games are customarily played. Informants knew that they were being recorded, but were unaware of the purpose. The quality of the recordings was affected by their informal setting but it is hypothesised that the performance of the speakers was minimally affected by the recording situation; the microphone was placed in a central position, speakers were not asked to speak into the microphone and the cassette recorder was left unmonitored during the session. Parts of each recording were played to a number of native speakers who were asked their impressions as to how 'interested' the speakers sounded. This was done in an unsystematic way with the sole aim of authenticating the data on the assumption that speakers who sound interested when asking questions, are motivated and are actually involved in a communicative act. A restricted corpus of the games where native judges claimed speakers to be most interested was selected for special attention. The main part of the restricted corpus consisted of a recording of a number of speakers who had a rapid rate of delivery and whose speech often overlapped. Of the group of speakers in the restricted corpus, seven were aged between 24 and 28, and two between 62 and 65. All were born and live in Palermo4.

1

This kind of game is otherwise known as "20 questions" or "What's my line?". 3 It is worth noting here that the yes-no questions considered in this study were exclusively of the information seeking type. Requests of confirmation were not included in this category. 4 Recordings were made of two speakers from another part of Sicily, but their intonation was regarded to be substantially different from the Palermo speakers; in this account, their speech has been ignored.

25

2 Transcription Initially, every recording in the restricted corpus was transcribed using a phonetic interlinear transcription method where stressed syllables were marked [•] and unstressed syllables [.]. A sample transcription was checked by two British intonologists. The working model proposed in chapter 2, section 4 is used to analyse these interlinear transcriptions into tone units consisting of, maximally, a prehead, head and nuclear unit5, and minimally a nuclear unit. The nuclear unit is preceded in the interlinear transcriptions by a broken line. Despite the discussions on pitch glides and jumps in 3.2.2 of chapter 2, challenging Crystal's assumption that there is no functional difference between the two, it was decided to do the initial transcription without marking glides except in cases of complex tones which usually involve a change of pitch direction on one syllable, or nuclear units of one syllable (i.e. where there was no tail). This decision was made for various reasons: there appeared to be phonetic gradience rather than a clear-cut distinction between glides and jumps (see chapter 2, section 3.2.3 on slope); segmental phonetic factors such as the length of the vowel and the degree of voicing on the surrounding consonants played an important role in determining whether there was a perceived glide, and, if so, to what extent. The head was not overtly marked; it was recognised as beginning on the first syllable transcribed with a heavy dot. This means that a prehead may consist only of unstressed syllables. This decision is intended to make transcription simpler and more consistent6.

3 Formal analysis of the core corpus The core corpus was analysed into 247 tone units. The categories needed for the description of this corpus are laid out below.

3.1 Nuclear tones The nuclear tones listed were described according to direction, beginning point and range. We consider simple and complex tones.

5

As discussed in chapter 2, the tail was not treated as independent of the nuclear syllable; the term 'nuclear unit' is used to designate the segmental domain of the nuclear tone. 6 Problems involved in making a distinction between heads and preheads were discussed in chapter 2.

26 3.1.1 Simple Tones Tone

Beginning point

Range

Examples

Frequency of occurrence

Fall

High (H)

Level

Rise

Narrow (N)

6%

Wide (W)

40%

Mid (M)

16%

Extra low (EL)

8%

High (H)

4%

Low (L)

20%

Low (L)

Narrow (N)

3 %7

The terms high, mid, low and extra low are illustrated by the following four points within the pitch range of a given speaker.

H

*

M

•

L

•

EL

•

The term 'low fall' has not been used, in order to avoid the analogy with Crystal's low fall in English which has, in fact, a mid starting point. For consistency, these tones are all classified according to their beginning pitch (within the speaker's range). The term 'narrow implies that less than 1/2 of the pitch range 8 is covered and the term 'wide' that the pitch movement stretches over more than 3/4 of it.

For this reason no

categories of narrow or wide mid falls or extra low falls are made. The few rises recorded were all narrow in range.

3.1.1.1 Falling tones The majority of tones in the corpus were falling (70%). Examples of high wide and high narrow nuclear tones are given below:

These figures do not sum to 100% because of rounding down in calculation. Hie speaker's pitch range is assessed auditorily.

27 1. Ma VIEne d'inVERno?

(But does he come in the Winter?) 2. EH ... di penSIEro?

(Er... mental?)

3.1.1.2

Level tones

A special note on the distribution of the level nuclear contour is in order. This contour is used in non-final hesitation cases. The duration of the syllable is usually increased relative to nonhesitation non-final nuclear tones. This is analogous to cases which Crystal claims are ambiguous: where one transcriber perceived a level nuclear tone, another did not perceive nuclearity. He states that the "main area of confusion is with drawled tones, though these usually lack the prominence of nuclear function, being lax, quietly articulated, and usually occurring on grammatical words." (1969:217). Some of the above features are common to Palermo Italian level contours, especially that they are drawled and accompanied by hesitation (pauses or false starts). An example of such a level nucleus is in the first part of the following: 3. E' un laVOro che - che PUO' anche non ESsere retriBUIto?

(Is it a job which could also be unpaid?)

In this case, the hesitation occurs on 'che' (which) and the level pitch and drawl on the modal verb 'puo" (can). Since the high level contours in the corpus were of this type, we conclude that the high level tone is not a full member of the nuclear tone set (see, again, discussion of level tones in Crystal: 1969). Low level tones occurred in two types of (intonational) context. (i) They occurred in cases such as the second tone unit transcribed below:

28 4. che LAvoRAva anche all'aPERto

Such a tone unit has been reanalysed as being subordinate in some way to the first. A similar case is reanalysed and discussed in section 3.3.2 on falling heads (type 2) in example 11 : "D prete per benedire la casa", where "il prete" constitutes one tone unit, and "per benedire la casa" a second, subordinate one9. (ii) They occurred in cases where the preceding pitch was high, such as in the first tone unit of example 4 above, and in the following example: 5. Che STA faCENdo?

(What's he doing?) It is arguable whether the syllable 'sta' is actually stressed, rather than just pitch prominent. In either case, the fact that the prenuclear pitch is high is what is of interest here.

3.1.1.3 Rising tones It appears that rising tones too have a defective distribution in the data analysed. Apart from the cases of rises uttered by speakers of local varieties outside Palermo, rises only occurred on words in which the final syllable was stressed, such as in the following example of a monosyllabic word: 6.

Gliel'ha DETto LUI

(Did hg say it?) There were no high rises in the restricted corpus, although there were two in the material recorded and analysed in less detail. One of these was as follows:

9 It is acknowledged that cases such as the latter have been analysed for English as simply a continuation of the nuclear contour.

29 7. PrevalenteMENte in citTÀ?

(predominantly in town?)

3.1.2 Complex tones The frequency of transcribed complex tones was very low (only 5 occurrences in the whole corpus). It was noted that there appeared to be a ce cases of transcribed rise-fall had fully voiced syllable onsets and/or diphthongs in the nuclear syllable, as in the following example: 8.

E' un attiviTA' be' retriBUIta?

(Is it a well paid activity?) One case of a fall-rise was recorded: 9. E1 un laVOro menTAle?

(Is it an intellectual job?) Note that the pitch immediately before the nucleus is very low.

3.2

Types of Prehead

Five types of prehead were listed, low (85%), mid (10%), high (2%), falling (1%) and rising (1%). Falling and rising preheads are not provided for in British English inventories 10 .

3.3

Types of Head

Nine types of head were described. The head is defined as ending on the syllable before the nuclear syllable and the transition between the two is not taken into account in the classification. The categories used are described and exemplified below.

This is the case in published work on British English known to the author.

30

3.3.1 Falling 1 This is a falling series of (rhythmically) stressed and unstressed syllables. In longer tone units there is a tendency towards a levelling out after the fall. 10. E' un laVOro che si SVOLge in miNIEra?

(Is it a job which is done in the mines?)

3.3.2 Falling 2 This is a sharp fall followed by a levelling out at a low pitch. 11. II PREte PER beneDIre la CAsa. •

•

• • • • • U •

(The priest t o bless the house.)

This type of head was originally listed and the argument was put forward that 'casa' had the perceptual prominence of a nucleus despite the absence of dynamic pitch. But it is analysed in chapter 7 as exemplifying "early focus", in which a major phrase consists of two minor tone units, the first ending on 'prete' and being the stronger of the two. Thus, the second phrase is in some way subordinate to the first, as in example 4 above.

3.3.3 Falling 3 This is a falling series of stressed syllables from which any unstressed syllables fall. 12. OcCORrono MEZzi di tras PORto?

(Is a means of transport necessary?)

3.3.4 Falling 4 This is a falling series of stressed syllables with any unstressed syllables continuing on the same pitch as the previous stressed syllable.

31 13. In T U T t e le STANze NO. • • • *

' i*

( N o t in all t h e rooms.)

3.3.5 Fallings This is a rising series of stressed syllables i descent.

14. Ci SOno tuoi aMlci che lo PRAticano?

(Are there any of your friends who do it?)

3.3.6 Low A low level sequence of syllables.

15. E per eSEMpio le SCARpe? . •

.j

(And shoes for example?)

3.3.7 High A high level sequence of syllables. 16. Che STA faCENdo? • •| •

| •

.

( W h a t ' s he doing?)

3.3.8 Mid A mid level sequence of syllables.

which unstressed syllables form a gradual

32 17. In Minima PARte. •

• •i I» •

(To a small extent.)

3.3.9 Rising

A gradually rising series of stressed and unstressed syllables. 18. Ma CU ciu FAFFAri? •

j• .

(Who makes us do it?)

This is the one case where the marking of a stressed syllable has been questioned in retrospect, as indicated in the alternative transcription below: 19. Ma CU ciu fafFAri? •

'

j> .

(Who makes us do it?)

The syllable 'faf is pitch prominent but not stressed. The pitch prominence has the effect of 'setting off the low pitch on the following syllable. It is also of interest that examples 16-19, all involving a head which ends higher that the beginning point of the nuclear tone are transcribed with a low level nucleus. Although there was no discussion in Grice (1984) of problems in distinguishing extra low narrow falls from low level nuclear tones, it is difficult to be certain that a consistent distinction can be maintained. What appears to be functionally more important here is the transition from a high pitch at the end of the head to a relatively low pitch on the nuclear syllable. More will be said about unstressed high syllables immediately before the nucleus in the following chapter, as it is a common pattern in standard Italian declarative utterances.

33

3.4

Frequency of occurrence of prenuclear contours

The frequency of occurrence in the data of the above illustrated heads is as follows: Falling 1 23% Falling 2 10% Falling 3 2% Falling 4 3% Falling 5 1% Low 18% Mid 7% Rising 4% In addition, 24% of the tone units had no head, 13% of which had a prehead and 11% no prenuclear tune at all.

4 Functional analysis of the core corpus 4.1 Questions and statements The initial hypothesis was that yes-no questions were distinguished from other categories of utterance, especially statements, by the nuclear tone alone; the high fall being used for questions and the mid or extra-low fall for statements. However, the situation was not as clear-cut as this: the mid falling tone was used in some yes-no questions as well as statements. In fact, the interpretation of an utterance as a question or statement appeared to depend on the prenuclear contour with which the mid tone was combined. The following table shows which utterance types were signalled by a number of combinations of prenuclear contours and a mid fall nuclear tone: Prenuclear contour

Utterance type

Falling 1

Yes-no Question

Low

Yes-no Question

Low prehead

Yes-no Question

Mid

Statement

Rising

Statement

34 It is clear from the outline of the tunes that the yes-no question forms involve an upward movement towards the beginning point of the nuclear unit; the end of the prenuclear tune must therefore be lower than the beginning point of the nuclear unit. In the case of the statement, the end of the head must be higher than (or the same as) the pitch at the start of the nuclear unit 11 . Thus, the functional contrast between questions and statements is being signalled by the pitch movement (or jump) on a part of the tone unit which has little or no status, being simply a transition between two of its components. The following table shows the combinations of nuclear tones and heads, classified according to their end pitch, and whether they are used in questions (+Q) or other utterance types(-Q) 12 . Type of Nuclear Tone Type of Head (end)

FALL high

LEVEL

high

FALL mid

LEVEL

low

RISE

low

FALL

xlow

narrow wide LOW

+Q

+Q

+Q

MID HIGH

-Q

-Q

-Q

+Q

-Q

-Q

-Q -Q

+Q

-Q -Q

-Q

-Q

The table gives additional support to the hypothesis that it is the rising transition which signals interrogation. If the height of the nuclear tone were specified in relation to the the pitch of a preceding syllable, as in Crystal (1969), Couper-Kuhlen (1986) and Cruttenden (1986) which were discussed in section 3.3 of chapter 2, yes-no questions could be said to have high falls, and statements low falls. There is an exception in the table, the rising nuclear tone, which can be considered a contextually determined phenomenon; words with final stress have a rise instead of a high fall in yes-no questions.

4.1.1 Complex tones Rising-falling complex tones involved a foil down to the beginning point of the nuclear syllable, having a rise in the first part of the nuclear tone. The falling-rising tone had a rising transition between the head and nuclear unit (i.e. a rise up to the beginning point). 11 This can be seen not just in the case of the mid fall nuclear tone, but as was noted above, and typically, in the case of extra-low falls. A group of 13 tone units are excluded from the count; these are all non-final elements of lists which will be discussed in 4.2 below.

35

4.2 Lists Thirteen contours were excluded from the analysis in 4.1 above. These were all non-final items in lists. They were analysed as follows:

Prenuclear tune

Nuclear tone high

Falling 5 head

mid

1

Mid head

1

Low head

4

Low prehead

3

2

No head/prehead These contours have the same structural description as yes-no question contours in terms of their nuclear and head tunes, as well as the transition between them. They were distinguished from yes-no questions in Grice (1984) by being transcribed as drawled. It is clear from further auditory analysis that there is a difference in timing of the pitch movement, the impression being that the rising transition begins earlier than in yes-no questions, but this difference is not easy to record within the British-style analysis carried out. In both earlier and later studies, the analysis of the nuclear tone as a high or mid fall was the same in yes-no questions and lists; it was either a rhythmic or a pitch-timing difference which was recorded.

5 Main observations and conclusion The main findings of the above analysis are as follows: 1) The nuclear tone of information seeking yes-no questions in this variety of Italian is typically a mid or high fall; such tones sometimes occur in statements too. However, what distinguishes yes-no questions from statements is the pitch at the end of the head (or prehead) in relation to the start of the nuclear fall: in the former it is lower and in the latter it is higher. It is therefore the transition between the head and the nuclear unit which signals interrogativity, entailing a rise or skip up. This contravenes the componentiality principle mentioned in chapter 1. 2) Non-final clauses (particularly in lists) may also have high or mid falls preceded by a low pitch; they are nonetheless distinguished from yes-no questions by intonational means. Although, when excised from context, non-final clauses are syntactically indistinguishable from yes-no questions, they do not always sound interrogative, despite the jump up in pitch.

36 The auditory impression of these contours is that the rise up to the start of the fall begins earlier. 3) A few non-interrogative contours contained high pitched unstressed syllables prior to the nuclear syllable which could not easily be described as belonging to a head - at this stage, little can be said about them as there were too few examples in the corpus, but they are discussed (in chapter 4 with regard to Standard Italian, and in chapter 7 with regard to Palermo Italian. 4) Yes-no questions with a final stressed syllable have what can be transcribed as a rising nucleus. 5) A number of interrogative contours can be transcribed as having a rise-fall nuclear tone; their nuclear syllable typically contained fully voiced onsets and/or diphthongs. The transition between head and nuclear unit in these cases was falling rather than rising, a problem to which a solution is suggested in chapter 7. 6) Crystal (1969), Cruttenden (1986) and Couper-Kuhlen (1986) all take the transition jump from the last unstressed syllable in the head to the nucleus as the criterion for classifying tones as high (jump up) or low (jump down), rather than the position in the speaker's pitch range. This approach allows the description of the [±interrogative] distinction as being manifested by a high fall in questions and a low fall in final clauses of statements. In a way, such an analysis extends the domain of the nucleus to the left: the syllable before the nuclear syllable is referred to in the classification of the nuclear tone, although its broad class, determined according to the direction of movement, is not affected (see the quote from Crystal below). However, non-final clauses and questions (both involving a jump up) cannot be distinguished by broad class alone. In addition, broad class analysis would require a rising monosyllabic nucleus (observation 4) to be described as a contextually determined allotonic variant of the high fall. This is counter-intuitive, given the consensus view as to the "basic division of nuclear tones into rising and falling" (Crystal: 1969:210). On the other hand, the rise-fall and fall distinction in 5 above is less problematic for the British-style approach, as the final pitch direction, which is considered to be tho most important, is the same. Thus, it can be seen that it is not possible to account for all the observed data satisfactorily using a British-style nucleus plus head approach; one is led into a series of inconsistencies in the analysis. Referring back to the two questions posed in the first paragraph of the introduction of this chapter, it can be seen that (i) The common factor in all interrogatives is the rise, either as a skip up to the nucleus or as a glide up within the nuclear syllable (as discussed in 4 and 5), and (ii) this form cannot be adequately described using a model of intonation based on the nuclear tone, at least not with the British tradition of analysis. Given the inappropriateness of the British school approach to analysing Palermo Italian intonation, it is reasonable to think that Italian analysts may have better equipment with which to perform the analysis. It is to these analysts that we turn in the next chapter.

IV. Auditory studies of Italian intonation

1 Introduction In this chapter, we shall discuss a number of auditory treatments of Italian intonation, with the expectation that auditory accounts by Italian specialists will capture the major meaningful distinctions within the intonation system they are describing. The studies dealt with in detail below all describe what is claimed to be Standard Italian accents. However, they do not all appear to be describing the same variety. In order to explain the lack of uniformity in the intonational forms described, we shall preface a consideration of these accounts with a discussion of the status of a so-called Standard in Italian.

1.1 Varieties of Italian Lepschy and Lepschy (1977:63) claim that Standard Italian not only 'does not exist in actual usage but is not even an ideal to which existing varieties strive to conform'. However, although there is no counterpart to British RP, certain regional varieties are considered to be more prestigious than others. As in other countries, the cultural or economic importance of a town or area is reflected in the prestige attributed to the accent associated with it. Since Italy's unification into a Nation State was, in European terms, relatively recent, Rome does not serve as its cultural and economic centre in the same way as London does for England. Since the first writings in what was called the Italian language were based on the dialect of Tuscany, a Tuscan accent has always had considerable prestige. However, the popular expression coined once Rome had become the seat of the Italian government (which occurred as late as 1870): 'Lingua toscana in bocca romana' (the Tuscan language spoken by Romans) was an indicator of the change in attitude towards the Florentine or Tuscan accent in favour of the accent of the new capital. Although the Tuscan origins of the language were still acknowledged, the expression reflected, according to Camilli (1965:154), the belief that the Romans spoke the literary language 'better' than the Tuscans. Milan and the 'industrial triangle' (Milan, Turin and Genoa) have since become Italy's financial and commercial centre. As a result, some social groups now consider the accent of Milan to be the most prestigious (see Galli de' Paratesi: 1985). In the literature on Italian intonation there have been a number of approaches to the issue of which Italian intonation to describe. Agard and Di Pietro (1965) and Fiorelli (1965) claim to describe a standard variety. Chapallaz (1979) bases her account of Italian intonation on the pronunciation of speakers from various places within a geographical area bounded by Rome to the South and Florence to the North. She thus does not refer to one local or even regional variety but makes generalisations about a number of varieties. Lepschy (1978) described his own accent, that of an educated speaker from Venice. Other studies have merely claimed to

38 investigate 'Italian' intonation without reference to any particular variety, describing the speech of individual Italian informants (Fogarasi:1975, Ames: 1969). A comparison of the various studies on Italian intonation should take into account that it is not always the same variety which is being described, even if they are all dealing with the Italian language as spoken by educated speakers. In this chapter we deal almost exclusively with accounts of what the respective authors describe as Standard Italian.

1.2 Auditory treatments of Italian intonation D'Eugenio (1982:235) wrote in the introductory paragraphs of his chapter on Italian intonation: 'The existing material on the subject looks almost insignificant if compared with the number of scholarly studies devoted to English intonation. To start with, there is no basic work... all we have is a series of articles and some brief sections in books dealing with Italian phonetics'. Even amongst these studies, there is no general consensus as to which theoretical approach is best suited to Italian. The two most frequently cited works, by Chapallaz and Agard and Di Pietro, are based upon entirely different approaches. Chapallaz (1960, 1962, 1964, 1979) describes whole tunes on the basis of the model of British English propounded by Armstrong and Ward (1926), whereas Agard and Di Pietro follow the American model of Trager and Smith (1951), subdividing the speaker's pitch range into four phonemic levels. D'Eugenio (1982) adopts an approach which combines componential aspects of the British-style analysis as described in chapter 2 (e.g. Crystal: 1969) with the levels approach of the American school (e.g. Trager and Smith:1951), and Canepari (1976, 1979, 1980, 1985) develops a system which draws heavily on the work of Halliday (1967). Common to most studies is the assumption that an investigation of Italian intonation should begin by examining the role of intonation in distinguishing yes-no questions from statements. However, the intonational form used to signal this functional contrast is not characterised in the same way in all accounts. We shall examine whether this is due to the fact that the models differ in the way they capture intonational forms, or to the fact that the forms differ from one variety to another. Lepschy and Lepschy (1977:62) have implied the latter when they refer to the fact that, in Italian, 'the use of intonational systems as part of the grammar' may vary across accents. There follows a survey of some of these studies along with discussion of the theoretical approaches employed in each account, and their appropriateness for describing the particular intonational phenomena discussed in chapter 3. In particular, we shall examine how these treatments characterise yes-no questions, statements and non-final list items, both in terms of the general shape of the intonation contours, as well as the exact location of distinctive pitch. The survey will take the following format: (1) a brief summary of the descriptive framework used to perform the analysis, (2) a description of the basic tunes described; in order to aid comparison of the forms across models (an interlinear interpretation will be provided of the patterns described), and (3) an investigation of exactly where distinctive pitch is said to occur, particularly in relation to the nuclear syllable.

39

2 Agard and Di Pietro Within the context of a contrastive study of the sound systems of 'Standard' Italian and 'American' English, Agard and Di Pietro (1965) provide a 'phonemic' account of Italian intonation. They first illustrate perceived pitch with a line above the orthographic text and then perform a phonemic analysis within the American structuralist tradition (Trager and Smith:1951, Pike:1945). In section 2.1, we shall give a short description of the framework within which they have worked, and in section 2.2 we shall describe the patterns they propose for polar (yes-no) questions, statements and non-final list items.

2.1 Descriptive framework Agard and Di Pietro use the same framework for describing English and Italian. They describe pitch patterns as a series of levels; low, middle, high and overhigh, symbolised in this case by the superscript numbers 1>2.3 and 4 respectively. These superscript numbers are placed at strategic positions, 'pitch points', in the phrase. They are as follows (with optional points in parenthesis): (INITIAL)

(PRECENTRAL)

CENTER

(PREFINAL)

FINAL

where the CENTER occurs on the syllable with "phrase stress" corresponding to what has been described in the previous chapters as the nuclear syllable. The pitch movement may be unidirectional (falling or rising) from the CENTER to the FINAL point, or it may be bidirectional (falling-rising or rising-falling); in the latter case, a PREFINAL point specifies the level at the pivot. After the FINAL point there is a 'TERMINAL CONTOUR' which may be rising (t), falling ( i ) or unchanged (—»). From the examples given, "unchanged" appears to mean that the pitch is levelled out (see discussion of levelling in chapter 2) as illustrated by the "phonemic" and "phonetic" transcriptions of the following phrase (1951:61): 1. / 2 That's

3

"all2 -*/

where the CENTER is level 3, the FINAL point level 2, and the TERMINAL CONTOUR is "unchanged". Most of the examples given show the TERMINAL CONTOUR following the direction of the change in pitch between the FINAL and what precedes it. However, Agard and Di Pietro do give examples where a semantic contrast is produced when the TERMINAL CONTOUR signals a reversal of direction, e.g. (1965:65):

40 /^come si 1 "chiama 1 i/ /^come si 1 "chiama^ t /

What is your name?1 (signalling no involvement) 'What is your name?' (signalling curiosity in relation to the speaker's personal experience)

where " is placed before the nuclear syllable, the number 1 before it indicates the CENTER, and 1 at the end of the word 'chiama' indicates the FINAL pitch point.

2.2

Basic Patterns

Agard and Di Pietro describe their two "basic patterns" as consisting of the following primary contours2: / l " l l / (CENTER 1, FINAL 1, falling TERMINAL) for statements and /3"3t/ (CENTER 3, FINAL 3, rising TERMINAL) for yes-no questions; these are BP I and BP II respectively. However, the way they are said to typically combine with precontours is of particular interest. When the CENTER of BP I is preceded a) by one or more unstressed syllables, the pattern is a i " i i / , b) by one or more stressed syllables, the pattern is /3 1" 14-/ and c) by one or more stressed syllables, which, in turn, are preceded by an unstressed syllable, the pattern is 12 3 When the CENTER of BP II is preceded by any material at all, regardless of stress patterns, it is on a mid pitch, the pattern being /2 3"3t/. Below are interlinear transcriptions of the two Basic Patterns with the following precontours: (a) unstressed INITIAL, (b) stressed INITIAL and (c) unstressed INITIAL and stressed PRECENTRAL. 2.

BPI (c)

(b)

(a) da

RO - m a

'

•

DE v o

par-TI-re

mi C H I A m o Gio-VAN-ni

4 3 2

1

(From Rome.)

(I m u s t l e a v e . )

( M y n a m e is G i o v a n n i . )

3 . B P II da

RO - m a

(From Rome?)

1

(c)

(b)

(a)

DE v o par-TI -

re

(Must I leave?)

mi C H I A m o Gio-VAN-ni

(Is m y n a m e G i o v a n n i ? )

Wh questions are described as having the same basic pitch contour as statements. 2 This is a term from Pike (1945) who divides the intonation contour into "precontour" (equivalent to the British prenuclear contour, i.e. head and prehead) and "primary contour" (equivalent to the British nucleus and tail components and what has been called "nuclear unit" in chapter 2 above).

41 According to Agard and di Pietro, our "unchanged" TERMINAL CONTOUR in BP II signals "more coming" (1965:66), and is typically used in non-final phrases. There is no specific mention of list items, although examples include lists of numbers. An example such as the following appears to be in a context in which the intonation pattern of non-final list elements could be used. 4. nel cinquant-OT-to 4 3

2 • 1

•

•

•

(in fifty eight)

2.3

Location of distinctive pitch

The fact that the same pitch contour is transcribed before and after the stressed syllable (i.e. on the CENTER and FINAL) implies that any pitch movement occurs either before it (a jump or glide up to or down to it) or after it (a glide or jump down from or up from it). Thus in BP I, generally used in statements, there is a fall before the nuclear syllable (from level two or three to level one), and the pitch on the nuclear syllable has no linguistically relevant pitch change, since both the CENTER and the FINAL are at level l 3 . In BP II, used in yes-no questions, there is a rise before the nuclear syllable (from level two to level three) and, again, no linguistically significant pitch change throughout the nuclear syllable (which is at level 3). The above observations are not simply a result of notational conventions; the Basic Pattern proposed for English statements does, by contrast, allow for a linguistically relevant pitch change on the nuclear syllable. The English BP I involves a fall from level 3 on the CENTER to level 1 on the FINAL point, represented as Agard and di Pietro do not make any provision for unstressed syllables to carry distinctive pitch unless they are at the boundary of a phrase, either at the beginning, in which case they can carry the pitch assigned to the INITIAL point, or at the end, in which case they carry the TERMINAL CONTOUR. In other positions, they remain on the same pitch level as the stressed syllable which they follow.

3 Chapallaz Chapallaz's work on Italian intonation has appeared in a series of articles (1960, 1962, 1964) and a chapter of her book on Italian pronunciation (1979, second edition: 1986). She does not

3

Since each level is treated as an international phoneme-like entity, any variation within level 1 would be considered equivalent to allophonic variation in the segmental-phonemic domain, and thus linguistically irrelevant.

42 explicitly set out to provide a contrastive study of Italian and English intonation, although her book is primarily aimed at an anglophone audience.

3.1 Descriptive framework Chapallaz describes phrase length tunes or "sense groups". Following the approach of Armstrong and Ward (1926, second edition: 1931), she does not subdivide these tunes into smaller units4. Since she considers tunes to be indivisible entities, she categorises them according to the shape of the whole pattern, rather than according to the final portion of the contour (as was the case in Agard and di Pietro's account). She makes no explicit reference to a nucleus or tonic, although her classification of tunes suggests that there is an implicit recognition of such a concept. In fact, Chapallaz points out that the "focal point of a sense group [is] usually the last part. This means that the final stressed syllable, under ordinary circumstances, is also the most important" (1979:162). Her classification of "stressed syllable" can be interpreted in a similar way to what Kingdon (1958) calls "fully stressed" or what Crystal refers to as "accented". Kingdon claims that the nucleus is associated with the "last fully stressed syllable of the group" (1958:6). Whereas Kingdon distinguishes between different degrees of stress and Crystal distinguishes stress and accent, Chapallaz has only a binary distinction for the purposes of her account of intonation; either a syllable is fully stressed, or it is not stressed at all. Chapallaz transcribes any syllables after the "focal point" with a light dot reserved for the transcription of unstressed syllables, even if they are lexically stressed. The focal point then becomes the last syllable of the group to be transcribed as stressed; it is therefore akin to the nucleus or tonic. Consequently, in the following exposition, reference will be made to the nucleus where Chapallaz had made reference to the final stressed syllable. As in English, the default position for the Italian nucleus is on the stressed syllable of the last lexical item in the group. However, there are cases where this rule can be overridden, as discussed in section 3.3 below.

3.2

Basic Patterns

Chapallaz describes three Basic Patterns, I (falling), II (falling-rising) and III (rising-falling). BP I is used in statements, BP II in yes-no questions, and BP III in non-final clauses, particularly in narrative speech, and on non-final list items. The general shape of each of the Basic Patterns is illustrated and exemplified below:

4

Armstrong and Ward explicitly declared that they did not consult previous studies on intonation, although they used the interlinear transcription system of Klinghardt (1923) with minor, typographically motivated changes.

43

eg Mi CHIAmo ROSsi.

(My name is Rossi.) 6. (ii) falling-rising

eg E'

fiNlto?

(Is it finished?) ( p i 92) 7. (iii) rising-falling

eg

Ha

suBlto..

(He underwent...) (p200)

Basic Patterns I and II have an alternative realisation which Chapallaz describes as having "breaks" where, in a gradually descending scale of syllables, one or sometimes more than one stressed syllable is raised to a higher pitch than the preceding syllable; the descent then continues from the raised syllable. This is illustrated by various renderings of the sentence 'Ho spesso sentito parlare di Lei' (I've often heard about you):

44 8.

(a) •t

• .

•• .

(b)

' • • t

•

ii.

(c)

—:

—

1

(P 185)

A fourth example of a tune with breaks appears to be of a different type: (d) NON mi

SENto

nemMEno

di

alZAre

un

BRACcio

stamatina.

(I don't even feel like lifting a finger this morning.) (p 186)

In this example, Chapallaz describes the word "nemmeno", which is in the middle of the utterance, as the most important. This utterance could, however, be analysed as two phrases, the first one being a BP Hi-type, ending with "nemmeno", and the second being a BP I type, ending with "stamattina". In this case there is a nuclear tone on "nemmeno" as well as on "braccio".

3.3

Location of distinctive pitch

Chapallaz's account of Italian intonation resembles Armstrong and Ward's account of English quite closely. Not only are BP I and II transcribed in almost the same way, they are used in similar contexts, the only exception being that yes-no questions can be spoken with BP I as well as BP II in English. The major difference is one of detail between English and Italian BP I; in English, the pitch on the nuclear syllable actually falls: "within the last stressed syllable, the pitch of the voice falls to a low level" (Armstrong and Ward,1926:4), whereas in Italian, there is usually a fall before the nuclear syllable and, on it, a "level or slightly falling pitch" (added emphasis) (Chapallaz, 1979:179). In BP II, the final rise usually begins on the nuclear syllable. However, a variant of BP II involves a fall on the nuclear syllable, followed by a rise. In Italian, this fall-rise ending tends

45 to have the effect of making questions sound "less peremptory" with a "gracious and even cajoling quality" (Chapallaz, 1979:198). Such a description closely resembles Armstrong and Ward's fall-rise ending to BP II, which, amongst other things, indicates "a wish to avoid appearing abrupt or dogmatic" and "a feeling of politeness" (Armstrong and Ward, 1926:56). Chapallaz's BP III does not have an English equivalent in Armstrong and Ward's account. Although Chapallaz describes the basic tune as "rising-falling", consisting of high pitch on the nuclear syllable, followed by slightly lower pitch on any unstressed syllables, the absence of unstressed syllables implies the absence of a falling element to the tune. Furthermore, she describes a variation of this tune which has the rise continuing on the postnuclear unstressed syllables. She claims that this is common in lists, an example of which is as follows: 9. SAN

FranCESco

(Saint Francis) (p 202)

This variation allows not only for contextually determined variation of the final pitch movement (no fall where no unstressed syllables are available to carry it), but also for context-free variation (unstressed syllables may form a rising sequence instead of a falling one). Variations on BP I and II preserve the final part of the tune, whereas the variation on BP III preserves the part of the tune up to the nuclear syllable, but not after it. It might be hypothesised that Chapallaz categorises this contour according to its functional similarity with unmodified BP III contours, rather than because its form is similar. On the other hand, it might be the case that Chapallaz observes a formal similarity which is not easily captured in terms of the nuclear tone approach; for instance, it might be that, for the categorisation of tunes into broad classes, the pitch before the nuclear syllable is as important as that on or after the nuclear syllable. A case where the pitch before the nuclear syllable is prominent, although it does not imply a change of broad class, is the following variant of BP I: 10. NON si

VEde

NULIa

(You can't see anything) (p 186)

where a single high unstressed syllable precedes the nuclear syllable. There is a similar variant of BP II. Chapallaz claims that the wide pitch interval between the high unstressed syllable and the nuclear syllable "gives added importance to the word to which the final stressed syllable [=nuclear syllable] belongs." (1979:186).

46

Chapallaz gives a number of examples of cases where the "focal point" of the utterance occurs before the last lexically stressed item. This can occur in statements or Wh-questions followed by a 'short group of parenthetical nature' in the form of unstressed syllables on a low level pitch. 11. Non

abBIAmo

MOLto TEMpo,

*

•

mi pare

V

(We don't have much time, it seems) ( p 1 8 4 )

She also gives an example of an early "focal point" in a Basic Pattern II tune, although she does not discuss it. However, the postnuclear material does not constitute a parenthetical group but rather a postposed subject: 12. NON ti anDAva il

lavoro?

(Didn't you feel like doing the work?) (p 192)

Here the postnuclear syllables rise gradually5and no stress is marked on the lexically stressed syllable "vo" of "lavoro". It is possible to produce both of the above examples with an intonationally marked boundary of some kind after the nucleus; i.e. they could be phrased as follows: [Non abbiamo molto tempo] [mi pare] [Non ti andava] [il lavoro]

Furthermore, the location of the "focal point" is largely predictable. There are also examples of early focus in the section on 'emphasis for contrast': 13. La

•

siGNOra mi

chiama.

V

(The lady is calling me.) (p 2 0 9 )

We could analyse the above two examples as having a low level tail in tune I and a rising tail in tune II.

47 14. La

•

siGNOra mi

- v .

chiama?

•

•

(Is the lady calling me?) (p 209)

In these examples, the postnuclear part does not constitute a syntactic postposition or parenthesis. The default location of the "focal point" would be on the final lexical item "chiama" rather than the earlier item "signora". In such contrastive cases, where the nucleus occurs early in a group, there is a greater pitch excursion on the nuclear syllable itself; the pitch falls actually on the nuclear syllable rather than down to it, as is the case in non-contrastive utterances 6 .

4

D'Eugenio

4.1 Descriptive framework D'Eugenio (1982) uses a similar phonetic interlinear transcription system to Chapallaz, although he claims to follow the analysis of Agard and Di Pietro, describing the basic elements of intonation as "a series of pitch levels and a terminal contour" (1982:232). However, in the introductory paragraphs of his chapter on intonation, he illustrates how a tone unit is subdivided into: (prehead), (head) 7 , (body), nucleus, (tail) and terminal contour, where the optional elements are in parenthesis. The major difference between this division of the tone unit and the British approach described in chapter 2 is the addition of the terminal contour following the tail. His account of Italian intonation is comparative, British English RP intonation being the object of comparison.

" This is an area where Palermo Italian differs from Northern Italian varieties, as will be shown in chapter 7. In Palermo Italian, the focussed item can be realised with a jump down in pitch to the nuclear syllable, even in cases of narrow (non-default) focus assignement. 7 DEugenio uses the terminology of Kingdon (1958), where "body" refers to what, within the British school, is now generally called the "head", and "head" refers to its first syllable. We shall continue to use the terms as they have been used in Chapter 2.

48

4.2

Basic Patterns

D'Eugenio discusses two basic tunes. Tune I consists of a falling head in which the syllables form a gradually descending scale or where each stressed syllable is on a slightly higher pitch than the (usually unstressed) syllable immediately preceding it 8 . D'Eugenio claims that, in tune II, the nucleus 'falls in a low pitch and then the voice rises up towards a mid level on the nuclear syllable if no unstressed syllable follows' (1982:241) and 'any unstressed syllable occurring after the nucleus is generally spoken on a higher pitch level than the syllable that precedes it.' (1982:241). However, he is later inconsistent with this position when he states that 'the voice pitch of the last fully stressed syllable falls in tune I and rises in tune n , as can be seen in the following examples (1982:236): 15. BP I PARto

BP I staSEra

(I'm leaving tonight)

DeSldera

qualCOsa?

(Do you want something?)

It appears that, in the claims made on page 241, he intended to say that the nucleus rises from low to mid and it is the prenuclear tune which falls to low, since otherwise the example would have to show the nucleus falling before it rises. According to D'Eugenio, non-final groups are spoken with tune II, e.g. (1982:241): 16. DoMEnica

SCORsa... (andammo al cinema) /

(Last Sunday...

we went to the cinema)

He describes a 'neutral terminal contour' where he claims that 'the pitch of the voice remains at a mid or high level or rises very slightly', (1982:247),although there are no examples of such a slightly rising terminal contour. The examples illustrated all consist of a level head on the same pitch as the following level nucleus, e.g. (1982:248):

This latter case is equivalent to Chapallaz's tune with "breaks".

49 17. GraDIsci un

.

bicCHIEre di... birra?

— . . .

•-> •

(Would you like a glass of ... beer?) This pattern occurs with unfinished phrases and is usually followed by a hesitation pause. It is also used in non-final list items, (p249) 9 : 18. Uno DUeTRE

QUATro

— . — . —

—

CINque... •

(One two three four five...) which, like other types of non-final groups, can also be spoken with tune II e.g. (p249) 19. DANte •

PeTRARca e •

BoCACcio * .

(Dante Petrarch and Boccaccio)

GerMAnia ^

(Germany)

4.3 Location of distinctive pitch According to D'Eugenio, the nucleus in BP I falls from mid to low; although it may be stepped down to as well, it does not necessarily have to be 10 . The distinctive pitch movement generally occurs actually on the nuclear syllable. The pitch in BP II generally involves a step down to the pitch on the nuclear syllable and a rise during it. D'Eugenio's account of Italian intonation differs little from his account of English in terms of the location of distinctive pitch.

9

This can be compared with Agard and Di Pietro's example (1965:66):

2"uno2-> 2"due2_> 2"tre2.> See the example of the last element of the list "Dante, Petrarca e Boccaccio" in 19 above.

50

5 Canepari Canepari's work on Italian intonation can be found in the introduction to his book on phonetics (1979), in his book on regional and standard Italian pronunciation (1980) and his book on intonation (1985). In these works, he discusses the intonation of Italian, both "Standard" and regional11, as well as that of other languages such as English.

5.1 Descriptive framework Following Halliday (1967), Canepari describes the tone group, which he calls intonia, as consisting of two main functional units or components: the pretonic (pretonia) and the tonic (tonia). Halliday's pretonic corresponds to what in chapter 2 was described as the prenuclear part of a tone unit which contains a head, that is, one of the following type: (prehead)

head

nucleus

(tail)

When a prehead occurs immediately before the nucleus, i.e. in a tone unit of type: prehead

nucleus

(tail)

hen the whole tune is classed as the tonic, including the prehead. Although Halliday acknowledges the tonic syllable, the equivalent of the nuclear syllable, he treats the tonic as a functionally indivisible entity in so far as it constitutes the domain of the tone. He explicitly points out, for instance, that there is no 'posttonic' element (1967:14) (which is equivalent to the tail discussed in chapter 2). Canepari, on the other hand, considers that the domain of the tone should be further subdivided; he refers to a postonica which has an identical description to the tail (all syllables after the tonic - or nuclear - syllable). In addition, he considers what precedes the tonic syllable (the tonia), which he calls protonica12 to be an independent functional unit too, even when there are stressed syllables in the pretonic part of the tone group (or, in other words, even when there is a head in the tone unit). Canepari's tone group can be represented as follows, where the major division of the tone group into pretonic and tonic is taken from Halliday, and the encircled part represents Canepari's own subdivision of the tonic:

11 Unfortunately, he does not describe Palermo Italian. He uses the term "pretónica" in later work and "protonica" in earlier work, referring to the same phenomenon.

51 tone group (intonîa)

pretonic (pretonìa)

tonic (tonîa)

postonica

5.2 Basic Patterns Canepari proposes three basic tunes, classified on functional grounds as (I) conclusive (used in neutral assertions), (II) interrogative (used in yes-no questions and requests for information), and (III) suspensive (used to signal that something interesting will follow, especially in narrative speech). These Basic Patterns are as follows: 20. BP I

RitorNIAmo doMEnica

(We'll return on Sunday.) 21. BP II RitorNIAmo doMEnica

(Shall we return on Sunday?)

52 22. BP III RitorNlamo doMEnica •

•

i •— •

(We'll return on Sunday,)

where the lines divide the speaker's pitch range up into three lingistically significant bands. Canepari claims that any pitch movement within one such band is not to be included into the phonological account.

5.3 Location of distinctive pitch Canepari maintains that the postónica (the part of the tonic which follows the tonic syllable) is important because it carries the significant pitch movement in the tone group - unless the tonic syllable is final in the phrase 13 . He maintains that the pretónica is subject to much regional variation, although within a given variety it tends to be the same for every tune. An exception is the accent of Catania, where the pretónica is high only when followed by a "supensive tonic" (exemplified in section 5.4). In varieties which have a high pretónica, the pretónica part is usually realised on one or more unaccented syllables before the tonic syllable (or tónica). However, in the absence of such unstressed syllables, the tonic syllable bears this pitch movement 14 . Such a readjustment is additional evidence for positing the pretónica as an integral part of an indivisible tonic, rather than as a separate functional unit within it (cf. arguments in chapter 2 against the nuclear syllable and tail as separate constituents). Nevertheless, the important contribution made by Canepari to the description of Italian intonation is the explicit acknowledgement that the pitch immediately prior to the tonic (or nuclear) syllable is linguistically significant. This will be further discussed in chapter 7, in the account of Palermo Italian intonation.

5.4 Canepari's account of "Sicilian Italian" In his account of Sicilian Italian, Canepari describes the accent of Catania rather than that of Palermo. He gives only the tonic part of the intonation contours for regional varieties, which are as follows for Catania Italian:

Such an argument resembles those put forward in favour of the tail. They have been discussed in chapter 2 and will not be discussed further here. This phenomenon is akin to that of Kingdon's homosyllabic preheads, which will be discussed in some detail in relation to English intonation in chapter 6.

53 23. BP I

BP II

•

• «

BP III

• ' .

These differ from Standard Italian, which he schematises as follows: 24. BP I

znmnz ••

BP II

• * »

BP III

• «•

The major differences are that the fall in BP I is sharper in Catania Italian than in Standard Italian, and that, in BP III, the tonic syllable is at a mid pitch in Catania Italian whereas it is high in Standard Italian. The interrogative contour is described in the same way. Although Canepari does not analyse Palermo Italian, the analysis in chapter 3 has presented the type of interlinear transcription of BP I, II and III which could be adapted to observe Canepari's three-banded division of the pitch range. Although, for the purposes of comparison, it would have been interesting to do this, it has not been done here; the decisions which would have had to have been made concerning the placement of pitch marks within appropriate bands would have required too much consideration of, or an unquestioned dependence on, the validity of such a tripartite division.

6 Fogarasi Fogarasi's account is discussed here, although it comprises but a short paper, because the interrogative contours he describes are radically different from those in any of the above accounts. Fogarasi describes what he calls "Italian intonation", although he does not give his source, whether it be his own competence of the language or that of a number of informants. The paper presents a contrastive study of Hungarian and Italian intonation; it is interesting to note that the intonation patterns used in Hungarian and Italian yes-no interrogatives are very similar, and are reproduced below: 25. HUNGARIAN Sze- re- lem?

ITALIAN Un a-MO- re?

(1975:10)

54

It might have appeared suspicious that the only radically different account of interrogative contours resembled the language with which Italian was being contrasted. However, it is the case that the rising-falling yes-no interrogative contours described resemble the Palermo Italian ones exemplified in chapter 3. Furthermore, the allophonic variant in the case of monosyllables is, in both languages, rising (cf. Varga:1984, Ladd:1981). Athough Canepari's description of local varieties of Italian from 21 different regions did not contain such a contour, Fogarasi provides evidence that the Palermo Italian contour is not an isolated phenomenon amongst these local varieties15.

7 Conclusions There is considerable variation regarding the different authors' descriptions of the Basic Patterns. We shall deal with each of the Basic Patterns in turn.

7.1 BP I Agard and di Pietro and Chapallaz both describe BP I as having a jump down to the nuclear syllable with a low level or slightly falling pitch on or from it. D'Eugenio describes a fall (from mid to low) on the nuclear syllable, and Canepari a fall from it. Common to all approaches appears to be the downward pitch movement.

7.2 BP II Agard and di Pietro present a jump up to a high level on the nuclear syllable, followed by a final glide up at the end of the phrase, Chapallaz and D'Eugenio describe a jump down to low, followed by a rise on or from the nuclear syllable, and Canepari describes a rise from the nuclear syllable. The common element to all approaches appears to be the upward pitch movement.

7.3 BP III Chapallaz details a jump up to the nuclear syllable, then a fall to mid on or from it, although the fall is not necessary: a variant of this tune involves a rise from the nuclear syllable. Canepari describes the same kind of rise then fall to mid, although he does not discuss the details of such a variant. Agard and di Pietro and D'Eugenio do not describe a pattern akin to Chapallaz's BP III, but describe a high level tone which is functionally similar.

An independent unpublished study by the present author found a similar interrogative pattern in Italian spoken in Crotone, Calabria. Since the completion of this study, a similar contour has been investigated in Barí Italian (Grice & Savino: 1994).

55

7.4 Distinctive pitch Agard and di Pietro and Canepari do not mark any pitch change through the nuclear syllable, implying that pitch change occurs before or after it. Agard and di Pietro account for pitch change after the nuclear syllable by means of the terminal contour, and Canepari does so by means of the "postonica" part of the tonic. Chapallaz and Canepari both claim that unstressed syllables immediately before the nuclear syllable may signal meaningful contrasts; Canepari incorporates this formally as the "protonica" part of the tonic, whereas Chapallaz considers such syllables as (albeit functionally significant) variations on the Basic Patterns. Chapallaz's tune III, which has a rising prenuclear and falling nuclear part has a variant in which the nuclear part is rising; in this case, it is the prenuclear part of the contour which is invariant.

7.5 Relation to analysis of Palermo Italian It has been shown in the above survey that a tune with a high pitch on the nuclear syllable followed by a fall can be placed in the same functional category as a tune with a high pitch with a successive fall. This was shown to be the case in Chapallaz's account of BP III and its contextually determined and context-free variants. This is similarto the case of a functionally equivalent rise and high fall in interrogative contours in Palermo Italian, even though the rise must be contextually determined in Palermo Italian. A description of a rising-falling contour, provided by Fogarasi, does have a similar contextually determined rising variant, even though it is unclear from his exposition exactly where the rising and falling elements occur in relation to the nuclear syllable. What Fogarasi's account and the one in chapter 3 have in common is the fact that in yes-no interrogatives they both have the rising-falling contour; all other varieties have a rising element without the fall. BP in, as described by Chapallaz and Canepari, is similar to the pattern used in both yesno questions (which, functionally, should have BP II) and non-final list elements (here Standard and Palermo Italian have a similar form-function relation). High unstressed syllables, as noted in chapter 3, are accounted for in a number of studies, and will be discussed further in relation to Palermo Italian in chapter 7.

7.6 Final remarks In this chapter, it has been seen that there is no proper consensus, either as to the form of the basic intonation patterns or how they should be characterised. Some commentators note the importance of immediately prenuclear pitch. In this respect, these treatments are more successful in accounting for the phenomena described in chapter 3. Canepari, for instance, accounts for unstressed prenuclear syllables within the "protonica". However, his framework cannot provide a principled way of accounting for all of the allophonic variants of the contour used in yes-no interrogatives in Palermo Italian.

56 It will be shown in the following chapters that an autosegmental theory of intonation can provide a mechanism for accounting for these allophonic variants, by formally separating boundary phenomena from tones associated with stressed (and a fortiori nuclear) syllables. Such a theory can also account for the position of distinctive pitch in the vicinity of the nuclear syllable, whether it is before, on or after it. In the following chapters, we shall investigate this theory further and adapt it for the description and analysis of Palermo Italian intonation.

V. Association in intonation models

In the next two chapters, the phenomena referred to as association and alignment will be examined in detail. Both deal with the synchronising, at different levels of abstraction, of a segmental phonemic component such as /pin/ with a specification for its intonation. At the most abstract level, vowels in a phoneme string may be linked to tones in a tonological string; each link is referred to as an association. The association comprises a link between the items in the two strings which leads towards their possible synchronisation at a realisational level. This synchronisation is referred to as alignment. In certain theories, associated elements are not always aligned with each other; as we shall see in more detail in chapter 6, an abstract link may undergo a process which causes one of the items to be shifted in time. In this chapter, we shall concentrate on association - the linking of items on separate tiers of a phonological representation - a concept which has its origins in the theory of autosegmental phonology. Section 1 will examine this theory in some detail, paying particular attention to the doctoral dissertation of one of its main protagonists: Goldsmith (1976). Section 2 will consider a selection of more recent studies that have considerably extended the concept of association. We shall refer to this work as post-autosegmental.

1 Autosegmental association The seminal work on autosegmental phonology is John Goldsmith's doctoral thesis, submitted in 1976 and published in its original state in 1979. Section 1.1 will lay out the basic principles of his thesis, examining its applications to different types of languages. In section 1.2, a distinction will be made between universal and language-specific phenomena. In section 1.3, we shall consider the way autosegmental phonology accounts for the lowering of one tone with respect to a previous one - the phenomenon referred to as downstep.

1.1 The theory In his dissertation, Goldsmith refers to a long tradition of treating melodic or intonational phenomena as distinct from items at a segmental phonemic level: hence the established use of the term "suprasegmental". He formalises this distinction by placing melodic and segmental phonemic items on separate tiers: a "tonal tier" for "tonemes" and a "CV tier" for "phonemes". The items in each of these tiers are of equal status and function as autonomous elements within their respective tiers.

58 African tone languages are the main object of his attention. However, he does make frequent references to English; for the purposes of this expository sketch, it will be mainly his English examples which we shall draw upon. In the citation form of the word "archipelago" [a:ki'pel@g@U] (Goldsmith's convention of using orthography where possible is retained in this example), a possible rendition would involve a fall from a high to a low pitch. This is represented as HL, a specification in the tonal tier, which is mapped onto vowels in the phonemic (or CV) tier by means of association

lines

thus: phoneme tier

tone tier

archipelago

W H

L

However, association lines are not drawn in an arbitrary way. Goldsmith makes a basic distinction between accentual and non-accentual languages. In non-accentual languages such as Mende, association is carried out by a process mapping one tone to one vowel, often from left to right, e.g. phoneme tier tone tier

m

ki

li

I L

I H

I L

In accentual languages, such as English, tonal melody is linked to the phonemic tier in a different way. One tonal element and one vowel are each marked with an asterisk or "star". Association is carried out in two stages: first an association line links the starred segments (Goldsmith uses the term "segment" to refer to an item on either of the tiers) as in 1(a): 1(a)

archipelago I H* L

then readjustments are made to the association so that the following conditions are met: la.

All tones are associated with at least one vowel,

b.

All vowels are associated with at least one tone.

2.

Association lines do not cross.

This would produce something like 1(b):

59 *

phoneme tier

archipelago

*

H L

tone tier

It is important that only minimal readjustments are made; i.e. that only as many association lines are added as are necessary to fulfil the conditions stated in 1 and 2 above, which are jointly referred to as the "well-formedness condition (WFC)" (Goldsmith 1979:27). Both the well-formedness condition as stated here and the constraint that only minimal adjustments should be made were originally intended to apply universally to all languages, whilst the star diacritic is specific to accentual languages. The star does not only indicate a starting point for the association rules; starred vowels and tones have a certain prominence which other, unstarred segments do not have. A starred association is accorded special status during the readjustment rules so as to preserve the prominence of the associated elements. To this aim, a starred element remains associated with only one item wherever possible; if other elements can be reassociated (or spread) instead of the starred item, then this is the preferred arrangement. This preference is exemplified below where the initial association is between V* i and T* \ as follows: 2(a)

V*,

T*i

V2

V3

T2

Of the following two readjustments : 2(b)

V*1

V2

V3

V2

V3

T*

2(c)

V*

T* 2(b) is preferred over 2(c), although the same number of association lines has been added in both cases. In 2(b) the pitch change resulting from the transition from the first to the second tone occurs between the starred vowel and another element, whereas in 2(c) it occurs between

60 two unstarred vowels. Since pitch movement lends perceptual prominence (see, inter alia, Bolinger:1958), it would make sense for the pitch movement to occur in the vicinity of the starred syllable. Goldsmith points out that one of the advantages of making the phoneme and tone tiers autonomous is that the citation form of words with a differing number of syllables may all have the same melodic representation, e.g. the representation H* L in 3(a) and (b) below: 3(a)

pi*n

N

H* L 3(b)

pe*ter

I I H* L In line with nuclear tone analyses, context dictates whether this fall involves a phonetic glide from high to low pitch on one syllable, or a skip (or glide) from one syllable to another. In Goldsmith's terminology, 'pin' has a contour tone and 'Peter' two level tones, but this is only when the autosegmental tiers have been fully associated; underlyingly, they both have the specification H*L on the tonal tier. Their functional equivalence is thus captured. One problem for English, brought up in Liberman (1975) and referred to in a footnote by Goldsmith, is the fact that a representation such as 4(a) especially elaborate

precautions

H L

where everything up to the stressed syllable in "precautions" is on a low pitch, represents the type of intonation contour produced in chanting, a particularly marked style of speech, rather than that used in normal conversation. He proposes a sparser specification, as in 4(b). 4(b) especially elaborate precautions

I

I

I

L

H

L

61 where there is some kind of interpolation between L and H (realised as a gradual rise in pitch). This gives a more accurate reflection of what is produced in conversational speech. Schematic versions of pitch contours for 4(a) and 4(b) might look something like the following: (a)

especially elaborate precautions

n_ L (b)

H

L

especially elaborate precautions

L

H

L

It could be argued, however, that (b) is not necessarily more natural than (a), but that they are simply two intonation contours with the same nuclear but distinct prenuclear configurations, (a) has a low and (b) has a rising head. The fact that the speaker has the opportunity to choose one or other contour indicates that (b) should be representable as distinct from (a). If the possibility of making this distinction is to be accepted, Goldsmith points out that the WFC would only be partially valid. A modified version includes parts l a and 2 only, as follows: la.

All tones are associated with at least one vowel.

2.

Association lines must not cross.

The use of the term "vowel" is shorthand for a segment which, in SPE (Chomsky and Halle: 1968) terms, has the feature [+syllabic]; segments without this feature are ignored in association rules. If, in principle, segments can be ignored, then parts la and lb of the WFC could be reformulated referring loosely to "tone-bearing units" (Goldsmith 1979:156) rather than to a specific feature a given segment may contain. The reformulation of part 1 might look like the following: la. lb.

All tones are associated with at least one tone-bearing unit, All tone-bearing units are associated with at least one tone.

This would allow a separate specification as to which items constitute tone bearing units in each language type: for accentual languages, only certain vowels are tone bearing units (i.e.

62 those which bear accent); for non-accentual languages, all vowels are tone bearing units. This interpretation means that the universal status of the WFC is maintained. As we shall see, the concept of accented syllables being tone-bearing units is crucial to the description of intonation within the frameworks discussed in section 2. As mentioned above, segments on the phonemic tier are specified in terms of feature matrices. The feature [±syllabic] is one of a number of features in a typical representation for [pin]. Goldsmith (1979:19) gives the following example: shorthand: matrices: +consonantal -nasal +labial -coronal

+syllabic -nasal -labial -coronal

+consonantal +nasal -labial +coronal

H and L are represented in the same way, shorthand: feature matrices:

H

L

[+ high] [-low]

[-high] [+low]

The use of the two binary features [±high] [±low], allows for the representation of a mid tone within this paradigm: M [+high] [+low] Goldsmith's proposal of one mid tone differs from that of Liberman (1975) which has two tones in the mid range: one with [+high] and [+low], the other with [-high] and [-low],

1.2 Universal and language-specific phenomena So far, we have seen how phonemic and tonal tiers are associated, first of all by an "initial association" and then by "readjustment rules". These rules are considered to be universal as they constitute ways of meeting the universal well-formedness condition. In addition, autosegmental theory allows for language-particular rules, permitting both the addition and deletion of association lines, the changing of a segment's features, and even,

63 within a given tier, the deletion of the segment itself. These rules are permitted within autosegmental theory, as long as adjustments are made at each step so as to ensure that the wellformedness condition is respected, e.g. a rule may cause one association line (A) to cross another (B); this can take place only if ( B ) is deleted by an adjustment rule in the next stage of the derivation. The adjustments are made to what is called the "chart" which consists of the two tiers and their related association lines. In autosegmental phonology, when an item is deleted from one tier, associated items on other tiers are not necessarily deleted along with it (in fact, they rarely are). This aspect of autosegmental behaviour is referred to as stability. Tonal stability is evident in tone languages where a morpheme involving a vowel with a particular tone is elided but the tone does not disappear. In the case below, taken from Lomongo, as described by Lovins (1971) and referred to by Goldsmith (1976:33), the first of two vowels spanning a word boundary is elided but its tone appears on the remaining vowel. "his"

"book"

balongo

bak a e

->

balong

a ka e

(elision rule specifies

I I I

A l l

I I

I l\ I I

deletion of element(s)

L

LH HH

L H

H LHHH

on phonemic tier)

H H

balong II

4

L H

H

a kae

(autosegmentally,

l\ I I

the vowel is delinked

LH H H

or disassociated from its tone)

->

balong II

a kae /l\ I I

L H H LH H H

(the tone is then reassociated with an adjacent vowel)

In the lexical representation, the final 'o' of 'balongo' is associated with a H tone. As the vowel 'o' is deleted, it is disassociated from the H tone. The H tone is then free to reassociate with another vowel (in fact, it must do, in order to fulfil the WFC). In this case, it is reassociated with the first vowel of the following word. A separate rule deletes the 'b' in 'bakae', but the process is irrelevant to the discussion here. The fact that H is a tone which, at some point in the derivation, is without an associated segment on the phonemic tier is not unusual in autosegmental phonology. This possibility has led to the use of the term "floating tone" for a tone which is not associated with a vowel. In certain African tone languages, there are morphemes which consist solely of tones. These tones are also 'floating'. They are dealt with in the same way as tones which are disassociated

64

during a derivation; they are associated with a vowel in an adjacent morpheme (see Clements and Ford (1979) for more on the theoretical status of floating tones).

1.3 Downstep One final aspect of autosegmental phonology to be considered here is the way it accounts for the lowering of one tone with respect to the preceding one: the phenomenon referred to, amongst other things, as downstep. Goldsmith looks at Igbo (1979:58ff) which appears to have three tones, H, L and M(id) (or Drop). He argues that pitch and tone should be kept apart in any description, pitch being almost as "superficial" as FO and tone being linguistically abstract. Accordingly, he claims that Igbo has 2 pitch registers, symbolised by a line for High tones (high register) and a line for Low tones (low register). Tones are placed on these register lines thus: 7. —H

high L

low

Common in African languages is the "pulling down" (1979:61) of the High register. The following type of rule governs this process: [+hi]

->

lowered

[+lo] ('M')

/

[+hi]

high register ('H'or'M')

i.e. a Mid tone is realised on a lowered high register if it is preceded by a Mid or High tone, otherwise a mid tone is realised on an unlowered high register. For instance, in a sequence: H

M

L

[+hi] [-lo]

[+hi] [+lo]

[-hi] [+lo]

the context for the rule (in italics) is met; the high register is therefore lowered as shown in the schema below: 8.

L-

65 If the sequence were MML rather than HML, the same register lowering would take place, and, in fact, the phonetic realisations of the respective sequences would be identical. The rule has an ambiguous status - it is not strictly phonological as it applies to the pitch register, which Goldsmith claims is very close to the phonetic representation. As we have seen, outside the scope of downstep, a degree of neutralisation takes place; that is, Mid tones are only phonetically distinct from High tones when they are immediately preceded by a tone with [+hi] in its specification. We shall see in the following sections that later models also have problems defining the level at which downstep operates.

2 Post-Autosegmental work on association A number of models have been developed since the work of Goldsmith, making use of its notational conventions and taking on board, to varying degrees, its theoretical basis. Common to all is the basic idea of tune and text as separate tiers although the association between them takes different forms. We shall refer mainly to work carried out on English and other so-called accentual languages (such as Swedish and Japanese). Accentual (as opposed to tonal) languages lend themselves to the type of sparse association suggested by Liberman (1975) where tones are associated with only certain vowels or syllables. Particular attention will be paid to work by Pierrehumbert, Beckman, Ladd, Hirst and Bruce. Where it is feasible, English examples will be given. Within these studies, the major divergences from Goldsmith are twofold: a. Tones are not simply associated with vowels: Pierrehumbert (1980) associates them with stressed (metrically strong) syllables, Hirst (1983) associates them with stress feet (and later to what he refers to as tonal units), and Bruce (1987) uses a hybrid approach. b. Tones may also be associated with the boundaries of higher-level constituents, such as the intonation phrase (roughly corresponding to a major tone group). This may be implicit (e.g. in the use of the term "boundary tone" in Pierrehumbert: 1980) or explicit (Hirst: 1983, Beckman and Pierrehumbert: 1986, Pierrehumbert and Beckman: 1988).

2.1 Association at different levels In the following sections, each model will be examined in relation to how it deals with (1) the association of tones with vowels, syllables or feet, the products of which are often referred to as pitch accents; and (2) the association of tones with higher-level constituents as mentioned in (b) above. We begin with a brief general word about these two categories of association.

66

2.1.1 Pitch Accents The term pitch accent is used by Bolinger (e.g. 1958, 1986) to refer to pitch configurations which accompany a prominent syllable. In the work described below, the configurations are further decomposed into combinations of H(igh) and L(ow) pitch levels of some kind. Bruce and Garding (1978) refer to High and Low turning points, Pierrehumbert refers to H(igh) and L(ow) tones, and Hirst(1983) and Ladd (1983) are closer to Goldsmith, referring to High and Low tones or peaks and valleys made up, inter alia, of thephonological features H(igh) and Uow). They all concur with Bolinger (1958) in their assumption that the pitch or FO contour carries an important message at these strategic points. However, the strong claim implied by Bolinger and incorporated into Pierrehumbert's formal account that each pitch accent implies a separate choice on the part of the speaker is challenged by Ladd (1986) who posits a hierarchical structuring which allows for the grouping of a number of pitch accents (e.g. into a " head" as described by O'Connor and Arnold, Crystal etc.). This will be further discussed in section 2.2.3.

2.1.2

Downstep

In addition, we shall examine how each of the models described account for the phonemenon of downstep referred to in 1.3 above.

2.1.3 Levels above the pitch accent Goldsmith's autosegmental phonology allowed tones to be associated only with vowels. It did not take higher constituents into account. Most of the work described below, on the other hand, allows for tones to be associated to the boundary of an intonation phrase. This is not particularly controversial if seen within the structuralist framework (e.g. Trager and Smith:1951, Pike:1945) which has "terminal tones" or "terminal contours" at the end of intonation phrases. The motivation for tones at the boundaries of other constituents is less universally accepted.

2.2 Specific models 2.2.1 Pierrehumbert (1980) (1) Pitch accents Janet Pierrehumbert's account of English allows for seven pitch accents consisting of H and L tones in combination with a star diacritic: H*. L*, H*+L, H+L*, L*+H, L+H* and H*+H

67 The starred tone is marked for association, in Goldsmith's sense, with a tone bearing unit. According to Pierrehumbert, the tone bearing unit is a syllable rather than a vowel; furthermore, it is metrically strong. Goldsmith's work referred to Chomsky and Halle's (1968) structure, where each segment was defined as a feature matrix and occurred in a linear string; each segment was defined separately. Since there was no bracketing, there was no direct way of representing syllable structure. The information had to be inferred from features such as [±syllabic] i.e. [+syllabic] segments consituted, in effect, syllable nuclei. Pierrehumbert has made use of later work which represents syllable structure in a more direct way. This enables her to define tone bearing units as certain types of syllable: metrically strong syllables. Metrical strength is determined according to a grid as in Liberman and Prince (1977). For Pierrehumbert, alignment is broadly fixed by association; i.e. a starred tone is realised within the time span occupied by its associated starred syllable. This can be exemplified as follows: the FO dip phonologically represented as L* in the pitch accent L*+H occurs during the portion of the signal corresponding to the metrically strong syllable with which it is associated. Thus, in 1 9. L*+H ... (other tonal material follows but is irrelevant here)

A

n

n

a

the FO is low on the first syllable and high after this. The exact position along the time axis at which the FO peak is reached depends on a number of factors, particularly how many tones and how much segmental material follow. These factors do not have any phonological status. The only difference in timing to which she accords phonological status is that between pairs such as L*+H and L+H*. These are considered to be entirely different pitch accents. Later work with Steele (Pierrehumbert and Steele 1989:19) indicates that, indeed, this distinction is perceived as categorical. 10.

»_ t 0

A

n

„ na

L*+H

1

L+H*

This is a close approximation to the FO contour taken from Pierrehumbert (1980:265).

68 In Pierrehumbert's model, the claim is made that tones represent targets directly in the FO contour without any need for an intermediate representation. Interpolation between these targets involves rules which are sensitive to the role of tones within the intonational structure (e.g. whether or not they are in a pitch accent, whether or not they are starred etc.). In her terms, these rules operate at the level of what she calls "phonetic realisation" which involves FO plotted against time rather than a quantity such as perceived pitch. An idea of the rules proposed can be gleaned from the following examples: between two successive H* accents of sufficient separation, there may be a dip ("sag") between the two peaks, the two H* peaks being, ceteris paribus, at the same height above a slightly declining baseline; in the sequences L * L * , L * H* and H* L, on the other hand, interpolation is monotonic. However, despite the emphasis on the acoustic-phonetic details, it should be pointed out that perceived pitch is mentioned where FO evidence requires corroboration, e.g. in the case of subordinate pitch accents which she calls echo accents (1980: ch 5, sect. 2; figs 5.10 5.12), or in certain cases of downstepped H* tones which do not show up as a clear obtrusion in the FO trace. She explicitly states about the word 'dots' upon which such a downstepped H* occurs: "To the ear, 'dots' is clearly not deaccented." (1980: fig 6.8). There is one case where a tone does not represent a target in the FO contour. This is the trailing L tone of the H * + L accent which is used in the representation of certain downstepped contours, as discussed below. (2) Downstep In Pierrehumbert's model, downstep is triggered by an alternating sequence of H and L tones. In certain cases, i.e. in H*+L, the L tone is present simply to contribute to the triggering sequence; it has no direct manifestation in the FO contour, i.e. it is not realised as a dip. Instead, it has the effect of lowering the FO value of a following High tone. She refers to this type of tone as floating because it is not directly associated to a tone bearing unit. To be exact, the trigger for downstep is as follows: In any H L H sequence containing a bitonal pitch accent, (e.g. H * + L H*, H* L+H*, or a combination of a bitonal pitch accent and a phrase accent) the second H has the FO value of the first, multiplied by a constant factor, k. The value of this constant factor (which must be between 0 and 1) is stipulated for the whole intonation phrase; the value used by Pierrehumbert is 0.6, (e.g. if the first H were at 200Hz, the second H would, ceteris paribus, be at 120Hz). The rule referring to the scaling of the second H is: Hi+i=kHj

where 0